To model complex real-world systems, such as traders in stock markets, or the dissemination of contagious diseases, graphon mean-field games (GMFG) have been proposed to model many agents. Despite the empirical success, our understanding of GMFG is limited. Popular algorithms such as mirror descent are deployed but remain unknown for their convergence properties. In this work, we give the first last-iterate convergence rate of mirror descent in regularized monotone GMFG. In tabular monotone GMFG with finite state and action spaces and under bandit feedback, we show a last-iterate convergence rate of O(T^{-1/4}). Moreover, when exact knowledge of costs and transitions is available, we improve this convergence rate to O(T^{-1}), matching the existing convergence rate observed in strongly convex games. In linear GMFG, our algorithm achieves a last-iterate convergence rate of O(T^{-1/5}). Finally, we verify the performance of the studied algorithms by empirically testing them against fictitious play in a variety of tasks.

In recent years, diffusion models have become the leading approach for distribution learning. This paper focuses on structure-preserving diffusion models (SPDM), a specific subset of diffusion processes tailored for distributions with inherent structures, such as group symmetries. We complement existing sufficient conditions for constructing SPDMs by proving complementary necessary ones. Additionally, we propose a new framework that considers the geometric structures affecting the diffusion process. Leveraging this framework, we design a structure-preserving bridge model that maintains alignment between the model’s endpoint couplings. Empirical evaluations on equivariant diffusion models demonstrate their effectiveness in learning symmetric distributions and modeling transitions between them. Experiments on real-world medical images confirm that our models preserve equivariance while maintaining high sample quality. We also showcase the practical utility of our framework by implementing an equivariant denoising diffusion bridge model, which achieves reliable equivariant image noise reduction and style transfer, irrespective of prior knowledge of image orientation.

Gradient-based multi-objective optimization (MOO) is essential in modern machine learning, with applications in e.g., multi-task learning, federated learning, algorithmic fairness and reinforcement learning. In this work, we first reveal some limitations of Pareto stationarity, a widely accepted first-order condition for Pareto optimality, in the presence of sparse function-variable structures. Next, to account for such sparsity, we propose a novel solution concept termed Refined Pareto Stationarity (RPS), which we prove is always sandwiched between Pareto optimality and Pareto stationarity. We give an efficient partitioning algorithm to automatically mine the function-variable dependency and substantially trim non-optimal Pareto stationary solutions. Then, we show that gradient-based descent algorithms in MOO can be enhanced with our refined partitioning. In particular, we propose Multiple Gradient Descent Algorithm with Refined Partition (RP-MGDA) as an example method that converges to RPS, while still enjoying a similar per-step complexity and convergence rate. Lastly, we validate our approach through experiments on both synthetic examples and realistic application scenarios where distinct function-variable dependency structures appear. Our results highlight the importance of exploiting function-variable structure in gradient-based MOO, and provide a seamless enhancement to existing approaches.

Adaptive moment estimation (Adam) is a cornerstone optimization algorithm in deep learning, widely recognized for its flexibility with adaptive learning rates and efficiency in handling large-scale data. However, despite its practical success, the theoretical understanding of Adam's convergence has been constrained by stringent assumptions, such as almost surely bounded stochastic gradients or uniformly bounded gradients, which are more restrictive than those typically required for analyzing stochastic gradient descent (SGD). In this paper, we introduce a novel and comprehensive framework for analyzing the convergence properties of Adam. This framework offers a versatile approach to establishing Adam's convergence. Specifically, we prove that Adam achieves asymptotic (last iterate sense) convergence in both the almost sure sense and the \(L_1\) sense under the relaxed assumptions typically used for SGD, namely \(L\)-smoothness and the ABC inequality. Meanwhile, under the same assumptions, we show that Adam attains non-asymptotic sample complexity bounds similar to those of SGD.

Recent diffusion-based generative models achieve remarkable results by training on massive datasets, yet this practice raises concerns about memorization and copyright infringement. A proposed remedy is to train exclusively on noisy data with potential copyright issues, ensuring the model never observes original content. However, through the lens of deconvolution theory, we show that although it is theoretically feasible to learn the data distribution from noisy samples, the practical challenge of collecting sufficient samples makes successful learning nearly unattainable. To overcome this limitation, we propose to pretrain the model with a small fraction of clean data to guide the deconvolution process. Combined with our Stochastic Forward--Backward Deconvolution (SFBD) method, we attain FID 6.31 on CIFAR-10 with just 4% clean images (and 3.58 with 10%). We also provide theoretical guarantees that SFBD learns the true data distribution. These results underscore the value of limited clean pretraining, or pretraining on similar datasets. Empirical studies further validate and enrich our findings.

We study the problem of no-regret learning algorithms for general monotone and smooth games and their last-iterate convergence properties. Specifically, we investigate the problem under bandit feedback and strongly uncoupled dynamics, which allows modular development of the multi-player system that applies to a wide range of real applications. We propose a mirror-descent-based algorithm, which converges in $O(T^{-1/4})$ and is also no-regret. The result is achieved by a dedicated use of two regularizations and the analysis of the fixed point thereof. The convergence rate is further improved to $O(T^{-1/2})$ in the case of strongly monotone games. Motivated by practical tasks where the game evolves over time, the algorithm is extended to time-varying monotone games. We provide the first non-asymptotic result in converging monotone games and give improved results for equilibrium tracking games.

Recently, deep multi-agent reinforcement learning (MARL) has demonstrated promising performance for solving challenging tasks, such as long-term dependencies and non-Markovian environments. Its success is partly attributed to conditioning policies on large fixed context length. However, such large fixed context lengths may lead to limited exploration efficiency and redundant information. In this paper, we propose a novel MARL framework to obtain adaptive and effective contextual information. Specifically, we design a central agent that dynamically optimizes context length via temporal gradient analysis, enhancing exploration to facilitate convergence to global optima in MARL. Furthermore, to enhance the adaptive optimization capability of the context length, we present an efficient input representation for the central agent, which effectively filters redundant information. By leveraging a Fourier-based low-frequency truncation method, we extract global temporal trends across decentralized agents, providing an effective and efficient representation of the MARL environment. Extensive experiments demonstrate that the proposed method achieves state-of-the-art (SOTA) performance on long-term dependency tasks, including PettingZoo, MiniGrid, Google Research Football (GRF), and StarCraft Multi-Agent Challenge v2 (SMACv2).

Diffusion-based purification (DBP) has become a cornerstone defense against adversarial examples (AEs), regarded as robust due to its use of diffusion models (DMs) that project AEs onto the natural data manifold. We refute this core claim, theoretically proving that gradient-based attacks effectively target the DM rather than the classifier, causing DBP's outputs to align with adversarial distributions. This prompts a reassessment of DBP's robustness, accrediting it two critical factors: inaccurate gradients and improper evaluation protocols that test only a single random purification of the AE. We show that when accounting for stochasticity and resubmission risk, DBP collapses. To support this, we introduce DiffBreak, the first reliable toolkit for differentiation through DBP, eliminating gradient mismatches that previously further inflated robustness estimates. We also analyze the current defense scheme used for DBP where classification relies on a single purification, pinpointing its inherent invalidity. We provide a statistically grounded majority-vote (MV) alternative that aggregates predictions across multiple purified copies, showing partial but meaningful robustness gain. We then propose a novel adaptation of an optimization method against deepfake watermarking, crafting systemic perturbations that defeat DBP even under MV, challenging DBP's viability.

Availability attacks, or unlearnable examples, are defensive techniques that allow data owners to modify their datasets in ways that prevent unauthorized machine learning models from learning effectively while maintaining the data's intended functionality. It has led to the release of popular black-box tools (e.g., APIs) for users to upload personal data and receive protected counterparts. In this work, we show that such black-box protections can be substantially compromised if a small set of unprotected in-distribution data is available. Specifically, we propose a novel threat model of protection leakage, where an adversary can (1) easily acquire (unprotected, protected) pairs by querying the black-box protections with a small unprotected dataset; and (2) train a diffusion bridge model to build a mapping between unprotected and protected data. This mapping, termed BridgePure, can effectively remove the protection from any previously unseen data within the same distribution. BridgePure demonstrates superior purification performance on classification and style mimicry tasks, exposing critical vulnerabilities in black-box data protection. We suggest that practitioners implement multi-level countermeasures to mitigate such risks.

In this work, we study potential games and Markov potential games under stochastic cost and bandit feedback. We propose a variant of the Frank-Wolfe algorithm with sufficient exploration and recursive gradient estimation, which provably converges to the Nash equilibrium while attaining sublinear regret for each individual player. Our algorithm simultaneously achieves a Nash regret and a regret bound of $O(T^{4/5})$ for potential games, which matches the best available result, without using additional projection steps. Through carefully balancing the reuse of past samples and exploration of new samples, we then extend the results to Markov potential games and improve the best available Nash regret from $O(T^{5/6})$ to $O(T^{4/5})$. Moreover, our algorithm requires no knowledge of the game, such as the distribution mismatch coefficient, which provides more flexibility in its practical implementation. Experimental results corroborate our theoretical findings and underscore the practical effectiveness of our method.

The family of probabilistic values, axiomatically-grounded in cooperative game theory, has recently received much attention in data valuation. However, it is often computationally expensive to compute exactly (exponential w.r.t. the number of data to valuate denoted by $n$). The existing generic estimator costs $O(n^2\log n)$ utility evaluations to achieve an $(\epsilon,\delta)$-approximation under the 2-norm, while faster estimators have been developed recently for special cases (e.g., empirically for the Shapley value and theoretically for the Banzhaf value). In this work, starting from the discovered connection between probabilistic values and least square regressions, we propose a Generic Estimator based on Least Squares (GELS) along with its variants that cost $O(n\log n)$ utility evaluations for many probabilistic values, largely extending the scope of this currently best complexity bound. Moreover, we show that each distributional value, proposed by Ghorbani et al. (2020) to alleviate the inconsistency of probabilistic values induced by using distinct databases, can also be cast as optimizing a similar least square regression. This observation leads to a theoretically-grounded framework TrELS (Training Estimators based on Least Squares) that can train estimators towards the specified distributional values without requiring any supervised signals. Particularly, the trained estimators are capable of predicting the corresponding distributional values for unseen data, largely saving the budgets required for running Monte-Carlo methods otherwise. Our experiments verify the faster convergence of GELS, and demonstrate the effectiveness of TrELS in learning distributional values. Our code is available at https://github.com/watml/fastpvalue.

Copyright infringement may occur when a generative model produces samples substantially similar to some copyrighted data that it had access to during the training phase. The notion of access usually refers to including copyrighted samples directly in the training dataset, which one may inspect to identify an infringement. We argue that such visual auditing largely overlooks a concealed copyright infringement, where one constructs a disguise that looks drastically different from the copyrighted sample yet still induces the effect of training Latent Diffusion Models on it. Such disguises only require indirect access to the copyrighted material and cannot be visually distinguished, thus easily circumventing the current auditing tools. In this paper, we provide a better understanding of such disguised copyright infringement by uncovering the disguises generation algorithm, the revelation of the disguises, and importantly, how to detect them to augment the existing toolbox. Additionally, we introduce a broader notion of acknowledgment for comprehending such indirect access.

High utility and rigorous data privacy are of the main goals of a federated learning (FL) system, which learns a model from the data distributed among some clients. The latter has been tried to achieve by using differential privacy in FL (DPFL). There is often heterogeneity in clients’ privacy requirements, and existing DPFL works either assume uniform privacy requirements for clients or are not applicable when server is not fully trusted (our setting). Furthermore, there is often heterogeneity in batch and/or dataset size of clients, which as shown, results in extra variation in the DP noise level across clients’ model updates. With these sources of heterogeneity, straightforward aggregation strategies, e.g., assigning clients’ aggregation weights proportional to their privacy parameters ($\epsilon$) will lead to lower utility. We propose Robust-HDP, which efficiently estimates the true noise level in clients’ model updates and reduces the noise-level in the aggregated model updates considerably. Robust-HDP improves utility and convergence speed, while being safe to the clients that may maliciously send falsified privacy parameter $\epsilon$ to server. Extensive experimental results on multiple datasets and our theoretical analysis confirm the effectiveness of Robust-HDP. Our code can be found here: https://github.com/Saber-mm/HDPFL.git

The concept of probabilistic values, such as Beta Shapley values and weighted Banzhaf values, has gained recent attention in applications like feature attribution and data valuation. However, exact computation of these values is often exponentially expensive, necessitating approximation techniques. Prior research has shown that the choice of probabilistic values significantly impacts downstream performance, with no universally superior option. Consequently, one may have to approximate multiple candidates and select the best-performing one. Although there have been many efforts to develop efficient estimators, none are intended to approximate all probabilistic values both simultaneously and efficiently. In this work, we embark on the first exploration of achieving this goal. Adhering to the principle of maximum sample reuse and avoiding amplifying factors, we propose a one-sample-fits-all framework parameterized by a sampling vector to approximate intermediate terms that can be converted to any probabilistic value. Leveraging the concept of $ (\epsilon, \delta) $-approximation, we theoretically identify a key formula that effectively determines the convergence rate of our framework. By optimizing the sampling vector using this formula, we obtain i) a one-for-all estimator that achieves the currently best time complexity for all probabilistic values on average, and ii) a faster generic estimator with the sampling vector optimally tuned for each probabilistic value. Particularly, our one-for-all estimator achieves the fastest convergence rate on Beta Shapley values, including the well-known Shapley value, both theoretically and empirically. Finally, we establish a connection between probabilistic values and the least square regression used in (regularized) datamodels, showing that our one-for-all estimator can solve a family of datamodels simultaneously. Our code is available at https://github.com/watml/one-for-all.

Indiscriminate data poisoning attacks aim to decrease a model's test accuracy by injecting a small amount of corrupted training data. Despite significant interest, existing attacks remain relatively ineffective against modern machine learning (ML) architectures. In this work, we introduce the notion of model poisoning reachability as a technical tool to explore the intrinsic limits of data poisoning attacks towards target parameters (i.e., model-targeted attacks). We derive an easily computable threshold to establish and quantify a surprising phase transition phenomenon among popular ML models: data poisoning attacks can achieve certain target parameters only when the poisoning ratio exceeds our threshold. Building on existing parameter corruption attacks and refining the Gradient Canceling attack, we perform extensive experiments to confirm our theoretical findings, test the predictability of our transition threshold, and significantly improve existing indiscriminate data poisoning baselines over a range of datasets and models. Our work highlights the critical role played by the poisoning ratio, and sheds new insights on existing empirical results, attacks and mitigation strategies in data poisoning.

The construction of adversarial examples usually requires the existence of soft or hard labels for each instance, with respect to which a loss gradient provides the signal for construction of the example. We show that for batch normalized deep image recognition architectures, intermediate latents that are produced after a batch normalization step by themselves suffice to produce adversarial examples using an intermediate loss solely utilizing angular deviations, without relying on any label. We motivate our loss through the geometry of batch normed representations and their concentration of norm on a hypersphere and distributional proximity to Gaussians. Our losses expand intermediate latent based attacks that usually require labels. The success of our method implies that leakage of intermediate representations may create a security breach for deployed models, which persists even when the model is transferred to downstream usage. Removal of batch norm weakens our attack, indicating it contributes to this vulnerability. Our attacks also succeed against LayerNorm empirically, thus being relevant for transformer architectures, most notably vision transformers which we analyze.

Differential privacy (DP) has emerged as a rigorous notion to quantify data privacy. Subsequently, Renyi differential privacy (RDP) becomes an alternative to the ordinary DP notion in both theoretical and empirical studies, for its convenient compositional rules and flexibility. However, most mechanisms with DP (RDP) guarantees are essentially based on randomizing a fixed, finite-dimensional vector output. In this work, following Hall et al. (2013) we further extend RDP to functional outputs, where the output space can be infinite-dimensional, and develop all necessary tools, *e.g.*, (subsampled) Gaussian mechanism, composition, and post-processing rules, to facilitate its practical adoption. As an illustration, we apply functional RDP (f-RDP) to functions in the reproducing kernel Hilbert space (RKHS) to develop a differentially private generative model (DPGM), where training can be interpreted as iteratively releasing loss functions (in an RKHS) with DP (RDP) guarantees. Empirically, the new training paradigm achieves a significant improvement in privacy-utility trade-off compared to existing alternatives, especially when $\epsilon=0.2$. Our code is available at https://github.com/dihjiang/DP-kernel.

Data valuation, a principled way to rank the importance of each training datum, has become increasingly important. However, existing value-based approaches (e.g., Shapley) are known to suffer from the stochasticity inherent in utility functions that render consistent and reliable ranking difficult. Recently, Wang and Jia (2023) proposed the noise-structure-agnostic framework to advocate the Banzhaf value for its robustness against such stochasticity as it achieves the largest safe margin among many alternatives. Surprisingly, our empirical study shows that the Banzhaf value is not always the most robust when compared with a broader family: weighted Banzhaf values. To analyze this scenario, we introduce the concept of Kronecker noise to parameterize stochasticity, through which we prove that the uniquely robust semi-value, which can be analytically derived from the underlying Kronecker noise, lies in the family of weighted Banzhaf values while minimizing the worst-case entropy. In addition, we adopt the maximum sample reuse principle to design an estimator to efficiently approximate weighted Banzhaf values, and show that it enjoys the best time complexity in terms of achieving an $(\epsilon, \delta)$-approximation. Our theory is verified under both synthetic and authentic noises. For the latter, we fit a Kronecker noise to the inherent stochasticity, which is then plugged in to generate the predicted most robust semi-value. Our study suggests that weighted Banzhaf values are promising when facing undue noises in data valuation.

In neural network binarization, BinaryConnect (BC) and its variants are considered the standard. These methods apply the sign function in their forward pass and their respective gradients are backpropagated to update the weights. However, the derivative of the sign function is zero whenever defined, which consequently freezes training. Therefore, implementations of BC (e.g., BNN) usually replace the derivative of sign in the backward computation with identity or other approximate gradient alternatives. Although such practice works well empirically, it is largely a heuristic or ``training trick.'' We aim at shedding some light on these training tricks from the optimization perspective. Building from existing theory on ProxConnect (PC, a generalization of BC), we (1) equip PC with different forward-backward quantizers and obtain ProxConnect++ (PC++) that includes existing binarization techniques as special cases; (2) derive a principled way to synthesize forward-backward quantizers with automatic theoretical guarantees; (3) illustrate our theory by proposing an enhanced binarization algorithm BNN++; (4) conduct image classification experiments on CNNs and vision transformers, and empirically verify that BNN++ generally achieves competitive results on binarizing these models.

Convergence to a saddle point for convex-concave functions has been studied for decades, while recent years has seen a surge of interest in non-convex (zero-sum) smooth games, motivated by their recent wide applications. It remains an intriguing research challenge how local optimal points are defined and which algorithm can converge to such points. An interesting concept is known as the local minimax point, which strongly correlates with the widely-known gradient descent ascent algorithm. This paper aims to provide a comprehensive analysis of local minimax points, such as their relation with other solution concepts and their optimality conditions. We find that local saddle points can be regarded as a special type of local minimax points, called uniformly local minimax points, under mild continuity assumptions. In (non-convex) quadratic games, we show that local minimax points are (in some sense) equivalent to global minimax points. Finally, we study the stability of gradient algorithms near local minimax points. Although gradient algorithms can converge to local/global minimax points in the non-degenerate case, they would often fail in general cases. This implies the necessity of either novel algorithms or concepts beyond saddle points and minimax points in non-convex smooth games.

Ji Xin, Raphael Tang, Yaoliang Yu, Jimmy Lin. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

The slow speed of BERT has motivated much research on accelerating its inference, and the early exiting idea has been proposed to make trade-offs between model quality and efficiency.This paper aims to address two weaknesses of previous work: (1) existing fine-tuning strategies for early exiting models fail to take full advantage of BERT; (2) methods to make exiting decisions are limited to classification tasks.We propose a more advanced fine-tuning strategy and a learning-toexit module that extends early exiting to tasks other than classification.Experiments demonstrate improved early exiting for BERT, with better trade-offs obtained by the proposed finetuning strategy, successful application to regression tasks, and the possibility to combine it with other acceleration methods.

Deep learning (DL) training is nondeterministic and such nondeterminism was shown to cause significant variance of model accuracy (up to 10.8%). Such variance may affect the validity of the comparison of newly proposed DL techniques with baselines. To ensure such validity, DL researchers and practitioners must replicate their experiments multiple times with identical settings to quantify the variance of the proposed approaches and baselines. Replicating and measuring DL variances reliably and efficiently is challenging and understudied.We propose a ready-to-deploy framework DEVIATE that (1) measures DL training variance of a DL model with minimal manual efforts, and (2) provides statistical tests of both accuracy and variance. Specifically, DEVIATE automatically analyzes the DL training code and extracts monitored important metrics (such as accuracy and loss). In addition, DEVIATE performs popular statistical tests and provides users with a report of statistical p-values and effect sizes along with various confidence levels when comparing to selected baselines.We demonstrate the effectiveness of DEVIATE by performing case studies with adversarial training. Specifically, for an adversarial training process that uses the Fast Gradient Signed Method to generate adversarial examples as the training data, DEVIATE measures a max difference of accuracy among 8 identical training runs with fixed random seeds to be up to 5.1%.Tool and demo links: https://github.com/lin-tan/DEVIATE

Hao Cheng, Xiaodong Liu, Lis Pereira, Yaoliang Yu, Jianfeng Gao. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

BinaryConnect (BC) and its many variations have become the de facto standard for neural network quantization. However, our understanding of the inner workings of BC is still quite limited. We attempt to close this gap in four different aspects: (a) we show that existing quantization algorithms, including post-training quantization, are surprisingly similar to each other; (b) we argue for proximal maps as a natural family of quantizers that is both easy to design and analyze; (c) we refine the observation that BC is a special case of dual averaging, which itself is a special case of the generalized conditional gradient algorithm; (d) consequently, we propose ProxConnect (PC) as a generalization of BC and we prove its convergence properties by exploiting the established connections. We conduct experiments on CIFAR-10 and ImageNet, and verify that PC achieves competitive performance.

Shift neural networks reduce computation complexity by removing expensive multiplication operations and quantizing continuous weights into low-bit discrete values, which are fast and energy-efficient compared to conventional neural networks. However, existing shift networks are sensitive to the weight initialization and yield a degraded performance caused by vanishing gradient and weight sign freezing problem. To address these issues, we propose S$^3$ re-parameterization, a novel technique for training low-bit shift networks. Our method decomposes a discrete parameter in a sign-sparse-shift 3-fold manner. This way, it efficiently learns a low-bit network with weight dynamics similar to full-precision networks and insensitive to weight initialization. Our proposed training method pushes the boundaries of shift neural networks and shows 3-bit shift networks compete with their full-precision counterparts in terms of top-1 accuracy on ImageNet.

Deep learning (DL) systems have been gaining popularity in critical tasks such as credit evaluation and crime prediction. Such systems demand fairness. Recent work shows that DL software implementations introduce variance: identical DL training runs (i.e., identical network, data, configuration, software, and hardware) with a fixed seed produce different models. Such variance could make DL models and networks violate fairness compliance laws, resulting in negative social impact. In this paper, we conduct the first empirical study to quantify the impact of software implementation on the fairness and its variance of DL systems. Our study of 22 mitigation techniques and five baselines reveals up to 12.6% fairness variance across identical training runs with identical seeds. In addition, most debiasing algorithms have a negative impact on the model such as reducing model accuracy, increasing fairness variance, or increasing accuracy variance. Our literature survey shows that while fairness is gaining popularity in artificial intelligence (AI) related conferences, only 34.4% of the papers use multiple identical training runs to evaluate their approach, raising concerns about their results’ validity. We call for better fairness evaluation and testing protocols to improve fairness and fairness variance of DL systems as well as DL research validity and reproducibility at large.

Out-of-distribution generalization is one of the key challenges when transferring a model from the lab to the real world. Existing efforts mostly focus on building invariant features among source and target domains. Based on invariant features, a high-performing classifier on source domains could hopefully behave equally well on a target domain. In other words, we hope the invariant features to be \emph{transferable}. However, in practice, there are no perfectly transferable features, and some algorithms seem to learn ``more transferable'' features than others. How can we understand and quantify such \emph{transferability}? In this paper, we formally define transferability that one can quantify and compute in domain generalization. We point out the difference and connection with common discrepancy measures between domains, such as total variation and Wasserstein distance. We then prove that our transferability can be estimated with enough samples and give a new upper bound for the target error based on our transferability. Empirically, we evaluate the transferability of the feature embeddings learned by existing algorithms for domain generalization. Surprisingly, we find that many algorithms are not quite learning transferable features, although few could still survive. In light of this, we propose a new algorithm for learning transferable features and test it over various benchmark datasets, including RotatedMNIST, PACS, Office-Home and WILDS-FMoW. Experimental results show that the proposed algorithm achieves consistent improvement over many state-of-the-art algorithms, corroborating our theoretical findings.

In natural language processing, a recently popular line of work explores how to best report the experimental results of neural networks.One exemplar publication, titled "Show Your Work: Improved Reporting of Experimental Results" (Dodge et al., 2019), advocates for reporting the expected validation effectiveness of the best-tuned model, with respect to the computational budget.In the present work, we critically examine this paper.As far as statistical generalizability is concerned, we find unspoken pitfalls and caveats with this approach.We analytically show that their estimator is biased and uses error-prone assumptions.We find that the estimator favors negative errors and yields poor bootstrapped confidence intervals.We derive an unbiased alternative and bolster our claims with empirical evidence from statistical simulation.Our codebase is at https://github.com/ castorini/meanmax.

Large-scale pre-trained language models such as BERT have brought significant improvements to NLP applications.However, they are also notorious for being slow in inference, which makes them difficult to deploy in realtime applications.We propose a simple but effective method, DeeBERT, to accelerate BERT inference.Our approach allows samples to exit earlier without passing through the entire model.Experiments show that DeeBERT is able to save up to ∼40% inference time with minimal degradation in model quality.Further analyses show different behaviors in the BERT transformer layers and also reveal their redundancy.Our work provides new ideas to efficiently apply deep transformer-based models to downstream tasks.

Generative adversarial networks (GANs) have become one of the most popular generative modeling techniques in machine learning. In this work, we study the statistical and robust properties of GANs for Gaussian mean estimation under Huber’s contamination model, where an epsilon proportion of training data may be arbitrarily corrupted. We prove that f-GAN, when equipped with appropriate discriminators, achieve optimal minimax rate, hence extending the recent result of Gao et al. (2019a). In contrast, we show that other GAN variants such as MMD-GAN (with Gaussian kernel) and W-GAN may fail to achieve minimax optimality. We further adapt f-GAN to the sparse and the unknown covariance settings. We perform numerical simulations to confirm our theoretical findings and reveal new insights on the importance of discriminators.

We investigate the ability of popular flow models to capture tail-properties of a target density by studying the increasing triangular maps used in these flow methods acting on a tractable source density. We show that the density quantile functions of the source and target density provide a precise characterization of the slope of transformation required to capture tails in a target density. We further show that any Lipschitz-continuous transport map acting on a source density will result in a density with similar tail properties as the source, highlighting the trade-off between the importance of choosing a complex source density and a sufficiently expressive transformation to capture desirable properties of a target density. Subsequently, we illustrate that flow models like Real-NVP, MAF, and Glow as implemented lack the ability to capture a distribution with non-Gaussian tails. We circumvent this problem by proposing tail-adaptive flows consisting of a source distribution that can be learned simultaneously with the triangular map to capture tail-properties of a target density. We perform several synthetic and real-world experiments to complement our theoretical findings.

Invariance (defined in a general sense) has been one of the most effective priors for representation learning. Direct factorization of parametric models is feasible only for a small range of invariances, while regularization approaches, despite improved generality, lead to nonconvex optimization. In this work, we develop a \emph{convex} representation learning algorithm for a variety of generalized invariances that can be modeled as semi-norms. Novel Euclidean embeddings are introduced for kernel representers in a semi-inner-product space, and approximation bounds are established. This allows invariant representations to be learned efficiently and effectively as confirmed in our experiments, along with accurate predictions.

Deep models, while being extremely flexible and accurate, are surprisingly vulnerable to “small, imperceptible” perturbations known as adversarial attacks. While the majority of existing attacks focus on measuring perturbations under the $\ell_p$ metric, Wasserstein distance, which takes geometry in pixel space into account, has long been known to be a suitable metric for measuring image quality and has recently risen as a compelling alternative to the $\ell_p$ metric in adversarial attacks. However, constructing an effective attack under the Wasserstein metric is computationally much more challenging and calls for better optimization algorithms. We address this gap in two ways: (a) we develop an exact yet efficient projection operator to enable a stronger projected gradient attack; (b) we show that the Frank-Wolfe method equipped with a suitable linear minimization oracle works extremely fast under Wasserstein constraints. Our algorithms not only converge faster but also generate much stronger attacks. For instance, we decrease the accuracy of a residual network on CIFAR-10 to $3.4%$ within a Wasserstein perturbation ball of radius $0.005$, in contrast to $65.6%$ using the previous Wasserstein attack based on an \emph{approximate} projection operator. Furthermore, employing our stronger attacks in adversarial training significantly improves the robustness of adversarially trained models.

We study unsupervised multilingual alignment, the problem of finding word-to-word translations between multiple languages without using any parallel data. One popular strategy is to reduce multilingual alignment to the much simplified bilingual setting, by picking one of the input languages as the pivot language that we transit through. However, it is well-known that transiting through a poorly chosen pivot language (such as English) may severely degrade the translation quality, since the assumed transitive relations among all pairs of languages may not be enforced in the training process. Instead of going through a rather arbitrarily chosen pivot language, we propose to use the Wasserstein barycenter as a more informative ``mean'' language: it encapsulates information from all languages and minimizes all pairwise transportation costs. We evaluate our method on standard benchmarks and demonstrate state-of-the-art performances.

Deep learning (DL) training algorithms utilize nondeterminism to improve models' accuracy and training efficiency. Hence, multiple identical training runs (e.g., identical training data, algorithm, and network) produce different models with different accuracies and training times. In addition to these algorithmic factors, DL libraries (e.g., TensorFlow and cuDNN) introduce additional variance (referred to as implementation-level variance) due to parallelism, optimization, and floating-point computation.

MOTIVATION: The mutations of cancers can encode the seeds of their own destruction, in the form of T-cell recognizable immunogenic peptides, also known as neoantigens. It is computationally challenging, however, to accurately prioritize the potential neoantigen candidates according to their ability of activating the T-cell immunoresponse, especially when the somatic mutations are abundant. Although a few neoantigen prioritization methods have been proposed to address this issue, advanced machine learning model that is specifically designed to tackle this problem is still lacking. Moreover, none of the existing methods considers the original DNA loci of the neoantigens in the perspective of 3D genome which may provide key information for inferring neoantigens' immunogenicity. RESULTS: In this study, we discovered that DNA loci of the immunopositive and immunonegative MHC-I neoantigens have distinct spatial distribution patterns across the genome. We therefore used the 3D genome information along with an ensemble pMHC-I coding strategy, and developed a group feature selection-based deep sparse neural network model (DNN-GFS) that is optimized for neoantigen prioritization. DNN-GFS demonstrated increased neoantigen prioritization power comparing to existing sequence-based approaches. We also developed a webserver named deepAntigen (http://yishi.sjtu.edu.cn/deepAntigen) that implements the DNN-GFS as well as other machine learning methods. We believe that this work provides a new perspective toward more accurate neoantigen prediction which eventually contribute to personalized cancer immunotherapy. AVAILABILITY AND IMPLEMENTATION: Data and implementation are available on webserver: http://yishi.sjtu.edu.cn/deepAntigen. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Function estimation under shape restrictions, such as convexity, has many practical applications and has drawn a lot of recent interests. In this work we argue that convexity, as a global property, is too strict and prone to outliers. Instead, we propose to use weakly convex functions as a simple alternative to quantify “approximate convexity”—a notion that is perhaps more relevant in practice. We prove that, unlike convex functions, weakly convex functions can exactly interpolate any finite dataset and they are universal approximators. Through regularizing the modulus of convexity, we show that weakly convex functions can be efficiently estimated both statistically and algorithmically, requiring minimal modifications to existing algorithms and theory for estimating convex functions. Our numerical experiments confirm the class of weakly convex functions as another competitive alternative for nonparametric estimation.

Ji Xin, Jimmy Lin, Yaoliang Yu. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Triangular map is a recent construct in probability theory that allows one to transform any source probability density function to any target density function. Based on triangular maps, we propose a general framework for high-dimensional density estimation, by specifying one-dimensional transformations (equivalently conditional densities) and appropriate conditioner networks. This framework (a) reveals the commonalities and differences of existing autoregressive and flow based methods, (b) allows a unified understanding of the limitations and representation power of these recent approaches and, (c) motivates us to uncover a new Sum-of-Squares (SOS) flow that is interpretable, universal, and easy to train. We perform several synthetic experiments on various density geometries to demonstrate the benefits (and short-comings) of such transformations. SOS flows achieve competitive results in simulations and several real-world datasets.

In distributional reinforcement learning (RL), the estimated distribution of value functions model both the parametric and intrinsic uncertainties. We propose a novel and efficient exploration method for deep RL that has two components. The first is a decaying schedule to suppress the intrinsic uncertainty. The second is an exploration bonus calculated from the upper quantiles of the learned distribution. In Atari 2600 games, our method achieves 483 % average gain across 49 games in cumulative rewards over QR-DQN. We also compared our algorithm with QR-DQN in a challenging 3D driving simulator (CARLA). Results show that our algorithm achieves nearoptimal safety rewards twice faster than QRDQN.

Novelty detection, a fundamental task in machine learning, has drawn a lot of recent attention due to its wide-ranging applications and the rise of neural approaches. In this work, we present a general framework for neural novelty detection that centers around a multivariate extension of the univariate quantile function. Our framework unifies and extends many classical and recent novelty detection algorithms, and opens the way to exploit recent advances in flow-based neural density estimation. We adapt the multiple gradient descent algorithm to obtain the first efficient end-to-end implementation of our framework that is free of tuning hyperparameters. Extensive experiments over a number of real datasets confirm the efficacy of our proposed method against state-of-the-art alternatives.

Numerous existing works have shown that, key to the efficiency of distributed machine learning (ML) is proper system and algorithm co-design: system design should be tailored to the unique mathematical properties of ML algorithms, and algorithms can be re-designed to better exploit the system architecture. While existing research has made attempts along this direction, many algorithmic and system properties that are characteristic of ML problems remain to be explored. Through an exploration of system-algorithm co-design, we build a new decentralized system Orpheus to support distributed training of a general class of ML models whose parameters are represented with large matrices. Training such models at scale is challenging: transmitting and checkpointing large matrices incur substantial network traffic and disk IO, which aggravates the inconsistency among parameter replicas. To cope with these challenges, Orpheus jointly exploits system and algorithm designs which (1) reduce the size and number of network messages for efficient communication, 2) incrementally checkpoint vectors for light-weight and fine-grained fault tolerance without blocking computation, 3) improve the consistency among parameter copies via periodic centralized synchronization and parameter-replicas rotation. As a result of these co-designs, communication and fault tolerance costs are linear to both matrix dimension and number of machines in the network, as opposed to being quadratic in existing systems. And the improved parameter consistency accelerates algorithmic convergence. Empirically, we show our system outperforms several existing baseline systems on training several representative large-scale ML models.

Latent prediction models, exemplified by multi-layer networks, employ hidden variables that automate abstract feature discovery. They typically pose nonconvex optimization problems and effective semi-definite programming (SDP) relaxations have been developed to enable global solutions (Aslan et al., 2014).However, these models rely on nonparametric training of layer-wise kernel representations, and are therefore restricted to transductive learning which slows down test prediction. In this paper, we develop a new inductive learning framework for parametric transfer functions using matching losses. The result for ReLU utilizes completely positive matrices, and the inductive learner not only delivers superior accuracy but also offers an order of magnitude speedup over SDP with constant approximation guarantees.

At their core, many unsupervised learning models provide a compact representation of homogeneous density mixtures, but their similarities and differences are not always clearly understood. In this work, we formally establish the relationships among latent tree graphical models (including special cases such as hidden Markov models and tensorial mixture models), hierarchical tensor formats and sum-product networks. Based on this connection, we then give a unified treatment of exponential separation in \emph{exact} representation size between deep mixture architectures and shallow ones. In contrast, for \emph{approximate} representation, we show that the conditional gradient algorithm can approximate any homogeneous mixture within $\epsilon$ accuracy by combining $O(1/\epsilon^2)$ ``shallow'' architectures, where the hidden constant may decrease (exponentially) with respect to the depth. Our experiments on both synthetic and real datasets confirm the benefits of depth in density estimation.

With ever growing data volume and model size, an error-tolerant, communication efficient, yet versatile distributed algorithm has become vital for the success of many large-scale machine learning applications. In this work we propose m-PAPG, an implementation of the flexible proximal gradient algorithm in model parallel systems equipped with the partially asynchronous communication protocol. The worker machines communicate asynchronously with a controlled staleness bound $s$ and operate at different frequencies. We characterize various convergence properties of m-PAPG: 1) Under a general non-smooth and non-convex setting, we prove that every limit point of the sequence generated by m-PAPG is a critical point of the objective function; 2) Under an error bound condition of convex objective functions, we prove that the optimality gap decays linearly for every $s$ steps; 3) Under the Kurdyka-Łojasiewicz inequality and a sufficient decrease assumption, we prove that the sequences generated by m-PAPG converge to the same critical point, provided that a proximal Lipschitz condition is satisfied.

We consider learning a distance metric in a weakly supervised setting where bags (or sets) of instances are labeled with bags of labels. A general approach is to formulate the problem as a Multiple Instance Learning (MIL) problem where the metric is learned so that the distances between instances inferred to be similar are smaller than the distances between instances inferred to be dissimilar. Classic approaches alternate the optimization over the learned metric and the assignment of similar instances. In this paper, we propose an efficient method that jointly learns the metric and the assignment of instances. In particular, our model is learned by solving an extension of k-means for MIL problems where instances are assigned to categories depending on annotations provided at bag-level. Our learning algorithm is much faster than existing metric learning methods for MIL problems and obtains state-of-the-art recognition performance in automated image annotation and instance classification for face identification.

The large model capacity of latent space models (LSMs) enables them to achieve great performance on various applications, but meanwhile renders LSMs to be prone to overfitting. Several recent studies investigate a new type of regularization approach, which encourages components in LSMs to be diverse, for the sake of alleviating overfitting. While they have shown promising empirical effectiveness, in theory why larger “diversity” results in less overfitting is still unclear. To bridge this gap, we propose a new diversity-promoting approach that is both theoretically analyzable and empirically effective. Specifically, we use near-orthogonality to characterize “diversity” and impose angular constraints (ACs) on the components of LSMs to promote diversity. A generalization error analysis shows that larger diversity results in smaller estimation error and larger approximation error. An efficient ADMM algorithm is developed to solve the constrained LSM problems. Experiments demonstrate that ACs improve generalization performance of LSMs and outperform other diversity-promoting approaches.

Classification problems with a large number of classes inevitably involve overlapping or similar classes. In such cases it seems reasonable to allow the learning algorithm to make mistakes on similar classes, as long as the true class is still among the top-k (say) predictions. Likewise, in applications such as search engine or ad display, we are allowed to present k predictions at a time and the customer would be satisfied as long as her interested prediction is included. Inspired by the recent work of [15], we propose a very generic, robust multiclass SVM formulation that directly aims at minimizing a weighted and truncated combination of the ordered prediction scores. Our method includes many previous works as special cases. Computationally, using the Jordan decomposition Lemma we show how to rewrite our objective as the difference of two convex functions, based on which we develop an efficient algorithm that allows incorporating many popular regularizers (such as the l2 and l1 norms). We conduct extensive experiments on four real large-scale visual category recognition datasets, and obtain very promising performances.

Adversarial machines, where a learner competes against an adversary, have regained much recent interest in machine learning. They are naturally in the form of saddle-point optimization, often with separable structure but sometimes also with unmanageably large dimension. In this work we show that adversarial prediction under multivariate losses can be solved much faster than they used to be. We first reduce the problem size exponentially by using appropriate sufficient statistics, and then we adapt the new stochastic variance-reduced algorithm of Balamurugan & Bach (2016) to allow any Bregman divergence. We prove that the same linear rate of convergence is retained and we show that for adversarial prediction using KL-divergence we can further achieve a speedup of #example times compared with the Euclidean alternative. We verify the theoretical findings through extensive experiments on two example applications: adversarial prediction and LPboosting.

Sparsity is an important modeling tool that expands the applicability of convex formulations for data analysis, however it also creates significant challenges for efficient algorithm design. In this paper we investigate the generalized conditional gradient (GCG) algorithm for solving sparse optimization problems--- demonstrating that, with some enhancements, it can provide a more efficient alternative to current state of the art approaches. After studying the convergence properties of GCG for general convex composite problems, we develop efficient methods for evaluating polar operators, a subroutine that is required in each GCG iteration. In particular, we show how the polar operator can be efficiently evaluated in learning low-rank matrices, instantiated with detailed examples on matrix completion and dictionary learning. A further improvement is achieved by interleaving GCG with fixed-rank local subspace optimization. A series of experiments on matrix completion, multi-class classification, and multi-view dictionary learning shows that the proposed method can significantly reduce the training cost of current alternatives.

Pooling plays an important role in generating a discriminative video representation. In this paper, we propose a new semantic pooling approach for challenging event analysis tasks (e.g., event detection, recognition, and recounting) in long untrimmed Internet videos, especially when only a few shots/segments are relevant to the event of interest while many other shots are irrelevant or even misleading. The commonly adopted pooling strategies aggregate the shots indifferently in one way or another, resulting in a great loss of information. Instead, in this work we first define a novel notion of semantic saliency that assesses the relevance of each shot with the event of interest. We then prioritize the shots according to their saliency scores since shots that are semantically more salient are expected to contribute more to the final event analysis. Next, we propose a new isotonic regularizer that is able to exploit the constructed semantic ordering information. The resulting nearly-isotonic support vector machine classifier exhibits higher discriminative power in event analysis tasks. Computationally, we develop an efficient implementation using the proximal gradient algorithm, and we prove new and closed-form proximal steps. We conduct extensive experiments on three real-world video datasets and achieve promising improvements.

Many real-world data arise naturally as tensors. Equipped with a low rank prior, learning algorithms can benefit from exploiting the rich dependency encoded in a tensor. Despite its prevalence in low-rank matrix learning, trace norm ceases to be tractable in tensors and therefore most existing works resort to matrix unfolding. Although some theoretical guarantees are available, these approaches may lose valuable structure information and are not scalable in general. To address this problem, we propose directly optimizing the tensor trace norm by approximating its dual spectral norm, and we show that the approximation bounds can be efficiently converted to the original problem via the generalized conditional gradient algorithm. The resulting approach is scalable to large datasets, and matches state-of-the-art recovery guarantees. Experimental results on tensor completion and multitask learning confirm the superiority of the proposed method.

With ever growing data volume and model size, an error-tolerant, communication efficient, yet versatile parallel algorithm has become a vital part for the success of many large-scale applications. In this work we propose mspg, an extension of the flexible proximal gradient algorithm to the model parallel and stale synchronous setting. The worker machines of mspg operate asynchronously as long as they are not too far apart, and they communicate efficiently through a dedicated parameter server. Theoretically, we provide a rigorous analysis of the various convergence properties of mspg, and a salient feature of our analysis is its seamless generality that allows both nonsmooth and nonconvex functions. Under mild conditions, we prove the whole iterate sequence of mspg converges to a critical point (which is optimal under convexity assumptions). We further provide an economical implementation of mspg, completely bypassing the need of keeping a local full model. We confirm our theoretical findings through numerical experiments.

Complex event detection on unconstrained Internet videos has seen much progress in recent years. However, state-of-the-art performance degrades dramatically when the number of positive training exemplars falls short. Since label acquisition is costly, laborious, and time-consuming, there is a real need to consider the much more challenging semantic event search problem, where no example video is given. In this paper, we present a state-of-the-art event search system without any example videos. Relying on the key observation that events (e.g. dog show) are usually compositions of multiple mid-level concepts (e.g. "dog," "theater," and "dog jumping"), we first train a skip-gram model to measure the relevance of each concept with the event of interest. The relevant concept classifiers then cast votes on the test videos but their reliability, due to lack of labeled training videos, has been largely unaddressed. We propose to combine the concept classifiers based on a principled estimate of their accuracy on the unlabeled test videos. A novel warping technique is proposed to improve the performance and an efficient highly-scalable algorithm is provided to quickly solve the resulting optimization. We conduct extensive experiments on the latest TRECVID MEDTest 2014, MEDTest 2013 and CCV datasets, and achieve state-of-the-art performances.

Clustering is the task of grouping a set of objects so that objects in the same cluster are more similar to each other than to those in other clusters. The crucial step in most clustering algorithms is to find an appropriate similarity metric, which is both challenging and problem-dependent. Supervised clustering approaches, which can exploit labeled clustered training data that share a common metric with the test set, have thus been proposed. Unfortunately, current metric learning approaches for supervised clustering do not scale to large or even medium-sized datasets. In this paper, we propose a new structured Mahalanobis Distance Metric Learning method for supervised clustering. We formulate our problem as an instance of large margin structured prediction and prove that it can be solved very efficiently in closed-form. The complexity of our method is (in most cases) linear in the size of the training dataset. We further reveal a striking similarity between our approach and multivariate linear regression. Experiments on both synthetic and real datasets confirm several orders of magnitude speedup while still achieving state-of-the-art performance.

High dimensional nonparametric regression is an inherently difficult problem with known lower bounds depending exponentially in dimension. A popular strategy to alleviate this curse of dimensionality has been to use additive models of \emphfirst order, which model the regression function as a sum of independent functions on each dimension. Though useful in controlling the variance of the estimate, such models are often too restrictive in practical settings. Between non-additive models which often have large variance and first order additive models which have large bias, there has been little work to exploit the trade-off in the middle via additive models of intermediate order. In this work, we propose salsa, which bridges this gap by allowing interactions between variables, but controls model capacity by limiting the order of interactions. salsas minimises the residual sum of squares with squared RKHS norm penalties. Algorithmically, it can be viewed as Kernel Ridge Regression with an additive kernel. When the regression function is additive, the excess risk is only polynomial in dimension. Using the Girard-Newton formulae, we efficiently sum over a combinatorial number of terms in the additive expansion. Via a comparison on 15 real datasets, we show that our method is competitive against 21 other alternatives.

Unsupervised learning of structured predictors has been a long standing pursuit in machine learning. Recently a conditional random field auto-encoder has been proposed in a two-layer setting, allowing latent structured representation to be automatically inferred. Aside from being nonconvex, it also requires the demanding inference of normalization. In this paper, we develop a convex relaxation of two-layer conditional model which captures latent structure and estimates model parameters, jointly and optimally. We further expand its applicability by resorting to a weaker form of inference---maximum a-posteriori. The flexibility of the model is demonstrated on two structures based on total unimodularity---graph matching and linear chain. Experimental results confirm the promise of the method.

Regularization has played a key role in deriving sensible estimators in high dimensional statistical inference. A substantial amount of recent works has argued for nonconvex regularizers in favor of their superior theoretical properties and excellent practical performances. In a different but analogous vein, nonconvex loss functions are promoted because of their robustness against "outliers". However, these nonconvex formulations are computationally more challenging, especially in the presence of nonsmoothness and non-separability. To address this issue, we propose a new proximal gradient meta-algorithm by rigorously extending the proximal average to the nonconvex setting. We formally prove its nice convergence properties, and illustrate its effectiveness on two applications: multi-task graph-guided fused lasso and robust support vector machines. Experiments demonstrate that our method compares favorably against other alternatives.

We aim to detect complex events in long Internet videos that may last for hours. A major challenge in this setting is that only a few shots in a long video are relevant to the event of interest while others are irrelevant or even misleading. Instead of indifferently pooling the shots, we first define a novel notion of semantic saliency that assesses the relevance of each shot with the event of interest. We then prioritize the shots according to their saliency scores since shots that are semantically more salient are expected to contribute more to the final event detector. Next, we propose a new isotonic regularizer that is able to exploit the semantic ordering information. The resulting nearly-isotonic SVM classifier exhibits higher discriminative power. Computationally, we develop an efficient implementation using the proximal gradient algorithm, and we prove new, closed-form proximal steps. We conduct extensive experiments on three real-world video datasets and confirm the effectiveness of the proposed approach.

How can one build a distributed framework that allows efficient deployment of a wide spectrum of modern advanced machine learning (ML) programs for industrial-scale problems using Big Models (100s of billions of parameters) on Big Data (terabytes or petabytes)- Contemporary parallelization strategies employ fine-grained operations and scheduling beyond the classic bulk-synchronous processing paradigm popularized by MapReduce, or even specialized operators relying on graphical representations of ML programs. The variety of approaches tends to pull systems and algorithms design in different directions, and it remains difficult to find a universal platform applicable to a wide range of different ML programs at scale. We propose a general-purpose framework that systematically addresses data- and model-parallel challenges in large-scale ML, by leveraging several fundamental properties underlying ML programs that make them different from conventional operation-centric programs: error tolerance, dynamic structure, and nonuniform convergence; all stem from the optimization-centric nature shared in ML programs' mathematical definitions, and the iterative-convergent behavior of their algorithmic solutions. These properties present unique opportunities for an integrative system design, built on bounded-latency network synchronization and dynamic load-balancing scheduling, which is efficient, programmable, and enjoys provable correctness guarantees. We demonstrate how such a design in light of ML-first principles leads to significant performance improvements versus well-known implementations of several ML programs, allowing them to run in much less time and at considerably larger model sizes, on modestly-sized computer clusters.

Topic models are effective probabilistic tools for processing large collections of unstructured data. With the exponential growth of modern industrial data, and consequentially also with our ambition to explore much bigger models, there is a real pressing need to significantly scale up topic modeling algorithms, which has been taken up in lots of previous works, culminating in the recent fast Markov chain Monte Carlo sampling algorithms in [10, 23] for the unsupervised latent Dirichlet allocation (LDA) formulations.

Multimedia event detection (MED) and multimedia event recounting (MER) are fundamental tasks in managing large amounts of unconstrained web videos, and have attracted a lot of attention in recent years. Most existing systems perform MER as a post-processing step on top of the MED results. In order to leverage the mutual benefits of the two tasks, we propose a joint framework that simultaneously detects high-level events and localizes the indicative concepts of the events. Our premise is that a good recounting algorithm should not only explain the detection result, but should also be able to assist detection in the first place. Coupled in a joint optimization framework, recounting improves detection by pruning irrelevant noisy concepts while detection directs recounting to the most discriminative evidences. To better utilize the powerful and interpretable semantic video representation, we segment each video into several shots and exploit the rich temporal structures at shot level. The consequent computational challenge is carefully addressed through a significant improvement of the current ADMM algorithm, which, after eliminating all inner loops and equipping novel closed-form solutions for all intermediate steps, enables us to efficiently process extremely large video corpora. We test the proposed method on the large scale TRECVID MEDTest 2014 and MEDTest 2013 datasets, and obtain very promising results for both MED and MER.

What is a systematic way to efficiently apply a wide spectrum of advanced ML programs to industrial scale problems, using Big Models (up to 100 s of billions of parameters) on Big Data (up to terabytes or petabytes)? Modern parallelization strategies employ fine-grained operations and scheduling beyond the classic bulk-synchronous processing paradigm popularized by MapReduce, or even specialized graph-based execution that relies on graph representations of ML programs. The variety of approaches tends to pull systems and algorithms design in different directions, and it remains difficult to find a universal platform applicable to a wide range of ML programs at scale. We propose a general-purpose framework, Petuum, that systematically addresses data- and model-parallel challenges in large-scale ML, by observing that many ML programs are fundamentally optimization-centric and admit error-tolerant, iterative-convergent algorithmic solutions. This presents unique opportunities for an integrative system design, such as bounded-error network synchronization and dynamic scheduling based on ML program structure. We demonstrate the efficacy of these system designs versus well-known implementations of modern ML algorithms, showing that Petuum allows ML programs to run in much less time and at considerably larger model sizes, even on modestly-sized compute clusters.

The representer theorem assures that kernel methods retain optimality under penalized empirical risk minimization. While a sufficient condition on the form of the regularizer guaranteeing the representer theorem has been known since the initial development of kernel methods, necessary conditions have only been investigated recently. In this paper we completely characterize the necessary and sufficient conditions on the regularizer that ensure the representer theorem holds. The results are surprisingly simple yet broaden the conditions where the representer theorem is known to hold. Extension to the matrix domain is also addressed.

The proximal map is the key step in gradient-type algorithms, which have become prevalent in large-scale high-dimensional problems. For simple functions this proximal map is available in closed-form while for more complicated functions it can become highly nontrivial. Motivated by the need of combining regularizers to simultaneously induce different types of structures, this paper initiates a systematic investigation of when the proximal map of a sum of functions decomposes into the composition of the proximal maps of the individual summands. We not only unify a few known results scattered in the literature but also discover several new decompositions obtained almost effortlessly from our theory.

It is a common practice to approximate complicated'' functions with more friendly ones. In large-scale machine learning applications, nonsmooth losses/regularizers that entail great computational challenges are usually approximated by smooth functions. We re-examine this powerful methodology and point out a nonsmooth approximation which simply pretends the linearity of the proximal map. The new approximation is justified using a recent convex analysis tool---proximal average, and yields a novel proximal gradient algorithm that is strictly better than the one based on smoothing, without incurring any extra overhead. Numerical experiments conducted on two important applications, overlapping group lasso and graph-guided fused lasso, corroborate the theoretical claims."

Structured sparse estimation has become an important technique in many areas of data analysis. Unfortunately, these estimators normally create computational difficulties that entail sophisticated algorithms. Our first contribution is to uncover a rich class of structured sparse regularizers whose polar operator can be evaluated efficiently. With such an operator, a simple conditional gradient method can then be developed that, when combined with smoothing and local optimization, significantly reduces training time vs. the state of the art. We also demonstrate a new reduction of polar to proximal maps that enables more efficient latent fused lasso.

Subspace learning seeks a low dimensional representation of data that enables accurate reconstruction. However, in many applications, data is obtained from multiple sources rather than a single source (e.g. an object might be viewed by cameras at different angles, or a document might consist of text and images). The conditional independence of separate sources imposes constraints on their shared latent representation, which, if respected, can improve the quality of the learned low dimensional representation. In this paper, we present a convex formulation of multi-view subspace learning that enforces conditional independence while reducing dimensionality. For this formulation, we develop an efficient algorithm that recovers an optimal data reconstruction by exploiting an implicit convex regularizer, then recovers the corresponding latent representation and reconstruction model, jointly and optimally. Experiments illustrate that the proposed method produces high quality results.

Despite the variety of robust regression methods that have been developed, current regression formulations are either NP-hard, or allow unbounded response to even a single leverage point. We present a general formulation for robust regression --Variational M-estimation--that unifies a number of robust regression methods while allowing a tractable approximation strategy. We develop an estimator that requires only polynomial-time, while achieving certain robustness and consistency guarantees. An experimental evaluation demonstrates the effectiveness of the new estimation approach compared to standard methods.

Sparse learning models typically combine a smooth loss with a nonsmooth penalty, such as trace norm. Although recent developments in sparse approximation have offered promising solution methods, current approaches either apply only to matrix-norm constrained problems or provide suboptimal convergence rates. In this paper, we propose a boosting method for regularized learning that guarantees $\epsilon$ accuracy within $O(1/\epsilon)$ iterations. Performance is further accelerated by interlacing boosting with fixed-rank local optimization---exploiting a simpler local objective than previous work. The proposed method yields state-of-the-art performance on large-scale problems. We also demonstrate an application to latent multiview learning for which we provide the first efficient weak-oracle.

Automated feature discovery is a fundamental problem in machine learning. Although classical feature discovery methods do not guarantee optimal solutions in general, it has been recently noted that certain subspace learning and sparse coding problems can be solved efficiently, provided the number of features is not restricted a priori. We provide an extended characterization of this optimality result and describe the nature of the solutions under an expanded set of practical contexts. In particular, we apply the framework to a semi-supervised learning problem, and demonstrate that feature discovery can co-occur with input reconstruction and supervised training while still admitting globally optimal solutions. A comparison to existing semi-supervised feature discovery methods shows improved generalization and efficiency.

Robust regression and classification are often thought to require non-convex loss functions that prevent scalable, global training. However, such a view neglects the possibility of reformulated training methods that can yield practically solvable alternatives. A natural way to make a loss function more robust to outliers is to truncate loss values that exceed a maximum threshold. We demonstrate that a relaxation of this form of ``loss clipping'' can be made globally solvable and applicable to any standard loss while guaranteeing robustness against outliers. We present a generic procedure that can be applied to standard loss functions and demonstrate improved robustness in regression and classification problems.

We prove that linear projections between distribution families with fixed first and second moments are surjective, regardless of dimension. We further extend this result to families that respect additional constraints, such as symmetry, unimodality and log-concavity. By combining our results with classic univariate inequalities, we provide new worst-case analyses for natural risk criteria arising in different fields. One discovery is that portfolio selection under the worst-case value-at-risk and conditional value-at-risk criteria yields identical portfolios.

Although 2D-based face recognition methods have made great progress in the past decades, there are also some unsolved problems such as PIE. Recently, more and more researchers have focused on 3D-based face recognition approaches. Among these techniques, facial feature point localization plays an important role in representing and matching 3D faces. In this paper, we present a novel feature point localization method on 3D faces combining global shape model and local surface model. Bezier surface is introduced to represent local structure of different feature points and global shape model is utilized to constrain the local search result. Experimental results based on comparison of our method and curvature analysis show the feasibility and efficiency of the new idea.