When deploying machine learning models in the real world, we often face the challenge of “unlearning” specific data points or subsets after training. Inspired by Domain-Adversarial Training of Neural Networks (DANN), we propose a novel algorithm,SURE, for targeted unlearning.SURE treats the process as a domain adaptation problem, where the “forget set” (data to be removed) and a validation set from the same distribution form two distinct domains. We train a domain classifier to discriminate between representations from the forget and validation sets.Using a gradient reversal strategy similar to DANN, we perform gradient updates to the representations to “fool” the domain classifier and thus obfuscate representations belonging to the forget set. Simultaneously, gradient descent is applied to the retain set (original training data minus the forget set) to preserve its classification performance. Unlike other unlearning approaches whose training objectives are built based on model outputs, SURE directly manipulates the representations.This is key to ensure robustness against a set of more powerful attacks than currently considered in the literature, that aim to detect which examples were unlearned through access to learned embeddings. Our thorough experiments reveal that SURE has a better unlearning quality to utility trade-off compared to other standard unlearning techniques for deep neural networks.

The use of deep neural networks in reinforcement learning (RL) often suffers from performance degradation as model size increases. While soft mixtures of experts (SoftMoEs) have recently shown promise in mitigating this issue for online RL, the reasons behind their effectiveness remain largely unknown. In this work we provide an in-depth analysis identifying the key factors driving this performance gain. We discover the surprising result that tokenizing the encoder output, rather than the use of multiple experts, is what is behind the efficacy of SoftMoEs. Indeed, we demonstrate that even with an appropriately scaled single expert, we are able to maintain the performance gains, largely thanks to tokenization.

Reinforcement learning from human feedback usually models preferences using a reward function that does not distinguish between people. We argue that this is unlikely to be a good design choice in contexts with high potential for disagreement, like in the training of large language models. We formalise and analyse the problem of learning a reward model that can be specialised to a user. Using the principle of empirical risk minimisation, we derive a probably approximately correct (PAC) bound showing the dependency of the approximation error on the number of training examples, as usual, and also on the number of human raters who provided feedback on them. Based on our theoretical findings, we discuss how to best collect pairwise preference data and argue that adaptive reward models should be beneficial when there is considerable disagreement among users. We also propose a concrete architecture for an adaptive reward model. Our approach leverages the observation that individual preferences can be captured as a linear combination of a set of general reward features. We show how to learn such features and subsequently use them to quickly adapt the reward model to a specific individual, even if their preferences are not reflected in the training data. We present experiments with large language models illustrating our theoretical results and comparing the proposed architecture with a non-adaptive baseline. Consistent with our analysis, the benefits provided by our model increase with the number of raters and the heterogeneity of their preferences. We also show that our model compares favourably to adaptive counterparts, including those performing in-context personalisation.

Information on trees at the individual level is crucial for monitoring forest ecosystems and planning forest management. Current monitoring methods involve ground measurements, requiring extensive cost, time and labour. Advances in drone remote sensing and computer vision offer great potential for mapping individual trees from aerial imagery at broad-scale. Large pre-trained vision models, such as the Segment Anything Model (SAM), represent a particularly compelling choice given limited labeled data. In this work, we compare methods leveraging SAM for the task of automatic tree crown instance segmentation in high resolution drone imagery in three use cases: 1) boreal plantations, 2) temperate forests, and 3) tropical forests. We also look into integrating elevation data into models, in the form of Digital Surface Model (DSM) information, which can readily be obtained at no additional cost from RGB drone imagery. We present BalSAM, a model leveraging SAM and DSM information, which shows potential over other methods, particularly in the context of plantations. We find that methods using SAM out-of-the-box do not outperform a custom Mask R-CNN, even with well-designed prompts. However, efficiently tuning SAM further and integrating DSM information are both promising avenues for tree crown instance segmentation models.

Large language models (LLMs) excel at few-shot in-context learning (ICL) -- learning from a few examples provided in context at inference, without any weight updates. Newly expanded context windows allow us to investigate ICL with hundreds or thousands of examples – the many-shot regime. Going from few-shot to many-shot, we observe significant performance gains across a wide variety of generative and discriminative tasks. While promising, many-shot ICL can be bottlenecked by the available amount of human-generated outputs. To mitigate this limitation, we explore two new settings: (1) "Reinforced ICL" that uses model-generated chain-of-thought rationales in place of human rationales, and (2) "Unsupervised ICL" where we remove rationales from the prompt altogether, and prompts the model only with domain-specific inputs. We find that both Reinforced and Unsupervised ICL can be quite effective in the many-shot regime, particularly on complex reasoning tasks. We demonstrate that, unlike few-shot learning, many-shot learning is effective at overriding pretraining biases, can learn high-dimensional functions with numerical inputs, and performs comparably to supervised fine-tuning. Finally, we reveal the limitations of next-token prediction loss as an indicator of downstream ICL performance.

Diffusion-relaxation MRI aims to extract quantitative measures that characterise microstructural tissue properties such as orientation, size, and shape, but long acquisition times are typically required. This work proposes a physics-informed learning framework to extract an optimal subset of diffusion-relaxation MRI measurements for enabling shorter acquisition times, predict non-measured signals, and estimate quantitative parameters. In vivo and synthetic brain 5D-Diffusion-T1-T2∗-weighted MRI data obtained from five healthy subjects were used for training and validation, and from a sixth participant for testing. One fully data-driven and two physics-informed machine learning methods were implemented and compared to two manual selection procedures and Cramér-Rao lower bound optimisation. The physics-informed approaches could identify measurement-subsets that yielded more consistently accurate parameter estimates in simulations than other approaches, with similar signal prediction error. Five-fold shorter protocols yielded error distributions of estimated quantitative parameters with very small effect sizes compared to estimates from the full protocol. Selected subsets commonly included a denser sampling of the shortest and longest inversion time, lowest echo time, and high b-value. The proposed framework combining machine learning and MRI physics offers a promising approach to develop shorter imaging protocols without compromising the quality of parameter estimates and signal predictions.

With the success of large language models (LLMs) of code and their use as code assistants (e.g. Codex used in GitHub Copilot), techniques for introducing domain-specific knowledge in the prompt design process become important. In this work, we propose a framework called Repo-Level Prompt Generator that learns to generate example-specific prompts using prompt proposals. The prompt proposals take context from the entire repository, thereby incorporating both the structure of the repository and the context from other relevant files (e.g. imports, parent class files). Our technique doesn't require any access to the weights of the LLM, making it applicable in cases where we only have black-box access to the LLM. We conduct experiments on the task of single-line code auto-completion using code repositories taken from Google Code archives. We demonstrate that an oracle constructed from our prompt proposals gives a relative improvement of 36% over Codex, showing the quality of these proposals. Further, we show that when we train a model to predict a prompt proposal, we can achieve significant performance gains over Codex and other baselines. We release our code, data, and trained checkpoints at: https://github.com/shrivastavadisha/repo_level_prompt_generation.

Biodiversity is declining at an unprecedented rate, impacting ecosystem services necessary to ensure food, water, and human health and well-being. Understanding the distribution of species and their habitats is crucial for conservation policy planning. However, traditional methods in ecology for species distribution models (SDMs) generally focus either on narrow sets of species or narrow geographical areas and there remain significant knowledge gaps about the distribution of species. A major reason for this is the limited availability of data traditionally used, due to the prohibitive amount of effort and expertise required for traditional field monitoring. The wide availability of remote sensing data and the growing adoption of citizen science tools to collect species observations data at low cost offer an opportunity for improving biodiversity monitoring and enabling the modelling of complex ecosystems. We introduce a novel task for mapping bird species to their habitats by predicting species encounter rates from satellite images, and present SatBird, a satellite dataset of locations in the USA with labels derived from presence-absence observation data from the citizen science database eBird, considering summer (breeding) and winter seasons. We also provide a dataset in Kenya representing low-data regimes. We additionally provide environmental data and species range maps for each location. We benchmark a set of baselines on our dataset, including SOTA models for remote sensing tasks. SatBird opens up possibilities for scalably modelling properties of ecosystems worldwide.

In image classification, it is common practice to train deep networks to extract a single feature vector per input image. Few-shot classification methods also mostly follow this trend. In this work, we depart from this established direction and instead propose to extract sets of feature vectors for each image. We argue that a set-based representation intrinsically builds a richer representation of images from the base classes, which can subsequently better transfer to the few-shot classes. To do so, we propose to adapt existing feature extractors to instead produce sets of feature vectors from images. Our approach, dubbed SetFeat, embeds shallow self-attention mechanisms inside existing encoder architectures. The attention modules are lightweight, and as such our method results in encoders that have approximately the same number of parameters as their original versions. During training and inference, a set-to-set matching metric is used to perform image classification. The effectiveness of our proposed architecture and metrics is demonstrated via thorough experiments on standard few-shot datasets-namely miniImageNet, tieredImageNet, and CUB-in both the 1- and 5-shot scenarios. In all cases but one, our method outperforms the state-of-the-art.

Transfer-learning methods aim to improve performance in a data-scarce target domain using a model pretrained on a data-rich source domain. A cost-efficient strategy, linear probing, involves freezing the source model and training a new classification head for the target domain. This strategy is outperformed by a more costly but state-of-the-art method – fine-tuning all parameters of the source model to the target domain – possibly because fine-tuning allows the model to leverage useful information from intermediate layers which is otherwise discarded by the later previously trained layers. We explore the hypothesis that these intermediate layers might be directly exploited. We propose a method, Head-to-Toe probing (Head2Toe), that selects features from all layers of the source model to train a classification head for the target-domain. In evaluations on the Visual Task Adaptation Benchmark-1k, Head2Toe matches performance obtained with fine-tuning on average while reducing training and storage cost hundred folds or more, but critically, for out-of-distribution transfer, Head2Toe outperforms fine-tuning. Code used in our experiments can be found in supplementary materials.

Although deep learning models have achieved state-of-the art performance on a number of vision tasks, generalization over high dimensional multi-modal data, and reliable predictive uncertainty estimation are still active areas of research. Bayesian approaches including Bayesian Neural Nets (BNNs) do not scale well to modern computer vision tasks, as they are difficult to train, and have poor generalization under dataset-shift. This motivates the need for effective ensembles which can generalize and give reliable uncertainty estimates. In this paper, we target the problem of generating effective ensembles of neural networks by encouraging diversity in prediction. We explicitly optimize a diversity inducing adversarial loss for learning the stochastic latent variables and thereby obtain diversity in the output predictions necessary for modeling multi-modal data. We evaluate our method on benchmark datasets: MNIST, CIFAR100, TinyImageNet and MIT Places 2, and compared to the most competitive baselines show significant improvements in classification accuracy, under a shift in the data distribution and in out-of-distribution detection. over 10% relative improvement in classification accuracy, over 5% relative improvement in generalizing under dataset shift, and over 5% better predictive uncertainty estimation as inferred by efficient out-of-distribution (OOD) detection.

We investigate the impact of aliasing on generalization in Deep Convolutional Networks and show that data augmentation schemes alone are unable to prevent it due to structural limitations in widely used architectures. Drawing insights from frequency analysis theory, we take a closer look at ResNet and EfficientNet architectures and review the trade-off between aliasing and information loss in each of their major components. We show how to mitigate aliasing by inserting non-trainable low-pass filters at key locations, particularly where networks lack the capacity to learn them. These simple architectural changes lead to substantial improvements in generalization on i.i.d. and even more on out-of-distribution conditions, such as image classification under natural corruptions on ImageNet-C [11] and few-shot learning on Meta-Dataset [26]. State-of-the art results are achieved on both datasets without introducing additional trainable parameters and using the default hyper-parameters of open source codebases.

Few-shot dataset generalization is a challenging variant of the well-studied few-shot classification problem where a diverse training set of several datasets is given, for the purpose of training an adaptable model that can then learn classes from \emph{new datasets} using only a few examples. To this end, we propose to utilize the diverse training set to construct a \emph{universal template}: a partial model that can define a wide array of dataset-specialized models, by plugging in appropriate components. For each new few-shot classification problem, our approach therefore only requires inferring a small number of parameters to insert into the universal template. We design a separate network that produces an initialization of those parameters for each given task, and we then fine-tune its proposed initialization via a few steps of gradient descent. Our approach is more parameter-efficient, scalable and adaptable compared to previous methods, and achieves the state-of-the-art on the challenging Meta-Dataset benchmark.

The goal of program synthesis from examples is to find a computer program that is consistent with a given set of input-output examples. Most learning-based approaches try to find a program that satisfies all examples at once. Our work, by contrast, considers an approach that breaks the problem into two stages: (a) find programs that satisfy only one example, and (b) leverage these per-example solutions to yield a program that satisfies all examples. We introduce the Cross Aggregator neural network module based on a multi-head attention mechanism that learns to combine the cues present in these per-example solutions to synthesize a global solution. Evaluation across programs of different lengths and under two different experimental settings reveal that when given the same time budget, our technique significantly improves the success rate over PCCoder [Zohar et. al 2018] and other ablation baselines.

One of the challenges in machine learning research is to ensure that presented and published results are sound and reliable. Reproducibility, that is obtaining similar results as presented in a paper or talk, using the same code and data (when available), is a necessary step to verify the reliability of research findings. Reproducibility is also an important step to promote open and accessible research, thereby allowing the scientific community to quickly integrate new findings and convert ideas to practice. Reproducibility also promotes the use of robust experimental workflows, which potentially reduce unintentional errors. In 2019, the Neural Information Processing Systems (NeurIPS) conference, the premier international conference for research in machine learning, introduced a reproducibility program, designed to improve the standards across the community for how we conduct, communicate, and evaluate machine learning research. The program contained three components: a code submission policy, a community-wide reproducibility challenge, and the inclusion of the Machine Learning Reproducibility checklist as part of the paper submission process. In this paper, we describe each of these components, how it was deployed, as well as what we were able to learn from this initiative.

Text-based games are a natural challenge domain for deep reinforcement learning algorithms. Their state and action spaces are combinatorially large, their reward function is sparse, and they are partially observable: the agent is informed of the consequences of its actions through textual feedback. In this paper we emphasize this latter point and consider the design of a deep reinforcement learning agent that can play from feedback alone. Our design recognizes and takes advantage of the structural characteristics of text-based games. We first propose a contextualisation mechanism, based on accumulated reward, which simplifies the learning problem and mitigates partial observability. We then study different methods that rely on the notion that most actions are ineffectual in any given situation, following Zahavy et al.'s idea of an admissible action. We evaluate these techniques in a series of text-based games of increasing difficulty based on the TextWorld framework, as well as the iconic game Zork. Empirically, we find that these techniques improve the performance of a baseline deep reinforcement learning agent applied to text-based games.

Experience replay is central to off-policy algorithms in deep reinforcement learning (RL), but there remain significant gaps in our understanding. We therefore present a systematic and extensive analysis of experience replay in Q-learning methods, focusing on two fundamental properties: the replay capacity and the ratio of learning updates to experience collected (replay ratio). Our additive and ablative studies upend conventional wisdom around experience replay {—} greater capacity is found to substantially increase the performance of certain algorithms, while leaving others unaffected. Counterintuitively we show that theoretically ungrounded, uncorrected n-step returns are uniquely beneficial while other techniques confer limited benefit for sifting through larger memory. Separately, by directly controlling the replay ratio we contextualize previous observations in the literature and empirically measure its importance across a variety of deep RL algorithms. Finally, we conclude by testing a set of hypotheses on the nature of these performance benefits.

Recent work suggests that Generative Adversarial Networks (GANs) benefit disproportionately from large mini-batch sizes. This finding is interesting but also discouraging – large batch sizes are slow and expensive to emulate on conventional hardware. Thus, it would be nice if there were some trick by which we could generate batches that were effectively big though small in practice. In this work, we propose such a trick, inspired by the use of Coreset-selection in active learning. When training a GAN, we draw a large batch of samples from the prior and then compress that batch using Coreset-selection. To create effectively large batches of real images, we create a cached dataset of Inception activations of each training image, randomly project them down to a smaller dimension, and then use Coreset-selection on those projected embeddings at training time. We conduct experiments showing that this technique substantially reduces training time and memory usage for modern GAN variants, that it reduces the fraction of dropped modes in a synthetic dataset, and that it helps us use GANs to reach a new state of the art in anomaly detection.

Graph-based neural network models are producing strong results in a number of domains, in part because graphs provide flexibility to encode domain knowledge in the form of relational structure (edges) between nodes in the graph. In practice, edges are used both to represent intrinsic structure (e.g., abstract syntax trees of programs) and more abstract relations that aid reasoning for a downstream task (e.g., results of relevant program analyses). In this work, we study the problem of learning to derive abstract relations from the intrinsic graph structure. Motivated by their power in program analyses, we consider relations defined by paths on the base graph accepted by a finite-state automaton. We show how to learn these relations end-to-end by relaxing the problem into learning finite-state automata policies on a graph-based POMDP and then training these policies using implicit differentiation. The result is a differentiable Graph Finite-State Automaton (GFSA) layer that adds a new edge type (expressed as a weighted adjacency matrix) to a base graph. We demonstrate that this layer can find shortcuts in grid-world graphs and reproduce simple static analyses on Python programs. Additionally, we combine the GFSA layer with a larger graph-based model trained end-to-end on the variable misuse program understanding task, and find that using the GFSA layer leads to better performance than using hand-engineered semantic edges or other baseline methods for adding learned edge types.

Graph neural networks (GNNs) have emerged as a powerful tool for learning software engineering tasks including code completion, bug finding, and program repair. They benefit from leveraging program structure like control flow graphs, but they are not well-suited to tasks like program execution that require far more sequential reasoning steps than number of GNN propagation steps. Recurrent neural networks (RNNs), on the other hand, are well-suited to long sequential chains of reasoning, but they do not naturally incorporate program structure and generally perform worse on the above tasks. Our aim is to achieve the best of both worlds, and we do so by introducing a novel GNN architecture, the Instruction Pointer Attention Graph Neural Networks (IPA-GNN), which achieves improved systematic generalization on the task of learning to execute programs using control flow graphs. The model arises by considering RNNs operating on program traces with branch decisions as latent variables. The IPA-GNN can be seen either as a continuous relaxation of the RNN model or as a GNN variant more tailored to execution. To test the models, we propose evaluating systematic generalization on learning to execute using control flow graphs, which tests sequential reasoning and use of program structure. More practically, we evaluate these models on the task of learning to execute partial programs, as might arise if using the model as a heuristic function in program synthesis. Results show that the IPA-GNN outperforms a variety of RNN and GNN baselines on both tasks.

We show that the sum of the implicit generator log-density $\log p_g$ of a GAN with the logit score of the discriminator defines an energy function which yields the true data density when the generator is imperfect but the discriminator is optimal, thus making it possible to improve on the typical generator (with implicit density $p_g$). To make that practical, we show that sampling from this modified density can be achieved by sampling in latent space according to an energy-based model induced by the sum of the latent prior log-density and the discriminator output score. This can be achieved by running a Langevin MCMC in latent space and then applying the generator function, which we call Discriminator Driven Latent Sampling~(DDLS). We show that DDLS is highly efficient compared to previous methods which work in the high-dimensional pixel space and can be applied to improve on previously trained GANs of many types. We evaluate DDLS on both synthetic and real-world datasets qualitatively and quantitatively. On CIFAR-10, DDLS substantially improves the Inception Score of an off-the-shelf pre-trained SN-GAN~\citep{sngan} from $8.22$ to $9.09$ which is even comparable to the class-conditional BigGAN~\citep{biggan} model. This achieves a new state-of-the-art in unconditional image synthesis setting without introducing extra parameters or additional training.

Convolutional Neural Networks (CNNs) have shown impressive performance in computer vision tasks such as image classification, detection, and segmentation. Moreover, recent work in Generative Adversarial Networks (GANs) has highlighted the importance of learning by progressively increasing the difficulty of a learning task Kerras et al. When learning a network from scratch, the information propagated within the network during the earlier stages of training can contain distortion artifacts due to noise which can be detrimental to training. In this paper, we propose an elegant curriculum-based scheme that smoothes the feature embedding of a CNN using anti-aliasing or low-pass filters. We propose to augment the training of CNNs by controlling the amount of high frequency information propagated within the CNNs as training progresses, by convolving the output of a CNN feature map of each layer with a Gaussian kernel. By decreasing the variance of the Gaussian kernel, we gradually increase the amount of high-frequency information available within the network for inference. As the amount of information in the feature maps increases during training, the network is able to progressively learn better representations of the data. Our proposed augmented training scheme significantly improves the performance of CNNs on various vision tasks without either adding additional trainable parameters or an auxiliary regularization objective. The generality of our method is demonstrated through empirical performance gains in CNN architectures across four different tasks: transfer learning, cross-task transfer learning, and generative models.

From the early days of computing, games have been important testbeds for studying how well machines can do sophisticated decision making. In recent years, machine learning has made dramatic advances with artificial agents reaching superhuman performance in challenge domains like Go, Atari, and some variants of poker. As with their predecessors of chess, checkers, and backgammon, these game domains have driven research by providing sophisticated yet well-defined challenges for artificial intelligence practitioners. We continue this tradition by proposing the game of Hanabi as a new challenge domain with novel problems that arise from its combination of purely cooperative gameplay with two to five players and imperfect information. In particular, we argue that Hanabi elevates reasoning about the beliefs and intentions of other agents to the foreground. We believe developing novel techniques for such theory of mind reasoning will not only be crucial for success in Hanabi, but also in broader collaborative efforts, especially those with human partners. To facilitate future research, we introduce the open-source Hanabi Learning Environment, propose an experimental framework for the research community to evaluate algorithmic advances, and assess the performance of current state-of-the-art techniques.

In this paper, we investigate the possibility of monitoring highway traffic based on videos whose frame rate is too low to accurately estimate motion features. The goal of the proposed method is to recognize traffic conditions instead of measuring them, as is usually the case. The main advantage of our approach comes from its ability to process low-frame-rate videos for which motion features cannot be estimated. Our method takes advantage of the highly redundant nature of traffic scenes that are pictured from a top-down perspective showing vehicles on a predominant asphalted road surrounded by background objects. Due to the limited variety of objects pictured in traffic scenes, our method gets to learn features that are specific to such images. With these features, our method is able to segment traffic images, classify traffic scenes, and estimate traffic density without requiring motion features. Different convolutional neural network models are proposed to segment traffic images in three different classes (Road, Car, and Background), classify traffic images into different categories (Empty, Fluid, Heavy, and Jam), and predict traffic density. We also propose a procedure to perform transfer learning of any of these models to new traffic scenes.

We introduce GuessWhat?!, a two-player guessing game as a testbed for research on the interplay of computer vision and dialogue systems. The goal of the game is to locate an unknown object in a rich image scene by asking a sequence of questions. Higher-level image understanding, like spatial reasoning and language grounding, is required to solve the proposed task. Our key contribution is the collection of a large-scale dataset consisting of 150K human-played games with a total of 800K visual question-answer pairs on 66K images. We explain our design decisions in collecting the dataset and introduce the oracle and questioner tasks that are associated with the two players of the game. We prototyped deep learning models to establish initial baselines of the introduced tasks.

Matrix factorization (MF) is one of the most popular techniques for product recommendation, but is known to suffer from serious cold-start problems. Item cold-start problems are particularly acute in settings such as Tweet recommendation where new items arrive continuously. In this paper, we present a meta-learning strategy to address item cold-start when new items arrive continuously. We propose two deep neural network architectures that implement our meta-learning strategy. The first architecture learns a linear classifier whose weights are determined by the item history while the second architecture learns a neural network whose biases are instead adjusted. We evaluate our techniques on the real-world problem of Tweet recommendation. On production data at Twitter, we demonstrate that our proposed techniques significantly beat the MF baseline and also outperform production models for Tweet recommendation.

It is commonly assumed that language refers to high-level visual concepts while leaving low-level visual processing unaffected. This view dominates the current literature in computational models for language-vision tasks, where visual and linguistic inputs are mostly processed independently before being fused into a single representation. In this paper, we deviate from this classic pipeline and propose to modulate the \emph{entire visual processing} by a linguistic input. Specifically, we introduce Conditional Batch Normalization (CBN) as an efficient mechanism to modulate convolutional feature maps by a linguistic embedding. We apply CBN to a pre-trained Residual Network (ResNet), leading to the MODulatEd ResNet (\MRN) architecture, and show that this significantly improves strong baselines on two visual question answering tasks. Our ablation study confirms that modulating from the early stages of the visual processing is beneficial.

Audio description (AD) provides linguistic descriptions of movies and allows visually impaired people to follow a movie along with their peers. Such descriptions are by design mainly visual and thus naturally form an interesting data source for computer vision and computational linguistics. In this work we propose a novel dataset which contains transcribed ADs, which are temporally aligned to full length movies. In addition we also collected and aligned movie scripts used in prior work and compare the two sources of descriptions. We introduce the Large Scale Movie Description Challenge (LSMDC) which contains a parallel corpus of 128,118 sentences aligned to video clips from 200 movies (around 150 h of video in total). The goal of the challenge is to automatically generate descriptions for the movie clips. First we characterize the dataset by benchmarking different approaches for generating video descriptions. Comparing ADs to scripts, we find that ADs are more visual and describe precisely what is shown rather than what should happen according to the scripts created prior to movie production. Furthermore, we present and compare the results of several teams who participated in the challenges organized in the context of two workshops at ICCV 2015 and ECCV 2016.

We present an approach based on feed-forward neural networks for learning the distribution over textual documents. This approach is inspired by the Neural Autoregressive Distribution Estimator (NADE) model which has been shown to be a good estimator of the distribution over discrete-valued high-dimensional vectors. In this paper, we present how NADE can successfully be adapted to textual data, retaining the property that sampling or computing the probability of an observation can be done exactly and efficiently. The approach can also be used to learn deep representations of documents that are competitive to those learned by alternative topic modeling approaches. Finally, we describe how the approach can be combined with a regular neural network N-gram model and substantially improve its performance, by making its learned representation sensitive to the larger, document-level context.

We introduce the Dynamic Capacity Network (DCN), a neural network that can adaptively assign its capacity across different portions of the input data. This is achieved by combining modules of two types: low-capacity sub-networks and high-capacity sub-networks. The low-capacity sub-networks are applied across most of the input, but also provide a guide to select a few portions of the input on which to apply the high-capacity sub-networks. The selection is made using a novel gradient-based attention mechanism, that efficiently identifies input regions for which the DCN’s output is most sensitive and to which we should devote more capacity. We focus our empirical evaluation on the Cluttered MNIST and SVHN image datasets. Our findings indicate that DCNs are able to drastically reduce the number of computations, compared to traditional convolutional neural networks, while maintaining similar or even better performance.

We present an autoencoder that leverages learned representations to better measure similarities in data space. By combining a variational autoencoder (VAE) with a generative adversarial network (GAN) we can use learned feature representations in the GAN discriminator as basis for the VAE reconstruction objective. Thereby, we replace element-wise errors with feature-wise errors to better capture the data distribution while offering invariance towards e.g. translation. We apply our method to images of faces and show that it outperforms VAEs with element-wise similarity measures in terms of visual fidelity. Moreover, we show that the method learns an embedding in which high-level abstract visual features (e.g. wearing glasses) can be modified using simple arithmetic.

We introduce a new representation learning approach for domain adaptation, in which data at training and test time come from similar but different distributions. Our approach is directly inspired by the theory on domain adaptation suggesting that, for effective domain transfer to be achieved, predictions must be made based on features that cannot discriminate between the training (source) and test (target) domains. The approach implements this idea in the context of neural network architectures that are trained on labeled data from the source domain and unlabeled data from the target domain (no labeled target-domain data is necessary). As the training progresses, the approach promotes the emergence of features that are (i) discriminative for the main learning task on the source domain and (ii) indiscriminate with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a new gradient reversal layer. The resulting augmented architecture can be trained using standard backpropagation and stochastic gradient descent, and can thus be implemented with little effort using any of the deep learning packages. We demonstrate the success of our approach for two distinct classification problems (document sentiment analysis and image classification), where state-of-the-art domain adaptation performance on standard benchmarks is achieved. We also validate the approach for descriptor learning task in the context of person re-identification application.

We present Neural Autoregressive Distribution Estimation (NADE) models, which are neural network architectures applied to the problem of unsupervised distribution and density estimation. They leverage the probability product rule and a weight sharing scheme inspired from restricted Boltzmann machines, to yield an estimator that is both tractable and has good generalization performance. We discuss how they achieve competitive performance in modeling both binary and real-valued observations. We also present how deep NADE models can be trained to be agnostic to the ordering of input dimensions used by the autoregressive product rule decomposition. Finally, we also show how to exploit the topological structure of pixels in images using a deep convolutional architecture for NADE.

Common representation learning (CRL), wherein different descriptions (or views) of the data are embedded in a common subspace, has been receiving a lot of attention recently. Two popular paradigms here are canonical correlation analysis (CCA)-based approaches and autoencoder (AE)-based approaches. CCA-based approaches learn a joint representation by maximizing correlation of the views when projected to the common subspace. AE-based methods learn a common representation by minimizing the error of reconstructing the two views. Each of these approaches has its own advantages and disadvantages. For example, while CCA-based approaches outperform AE-based approaches for the task of transfer learning, they are not as scalable as the latter. In this work, we propose an AE-based approach, correlational neural network (CorrNet), that explicitly maximizes correlation among the views when projected to the common subspace. Through a series of experiments, we demonstrate that the proposed CorrNet is better than AE and CCA with respect to its ability to learn correlated common representations. We employ CorrNet for several cross-language tasks and show that the representations learned using it perform better than the ones learned using other state-of-the-art approaches.

We present a mathematical construction for the restricted Boltzmann machine (RBM) that does not require specifying the number of hidden units. In fact, the hidden layer size is adaptive and can grow during training. This is obtained by first extending the RBM to be sensitive to the ordering of its hidden units. Then, with a carefully chosen definition of the energy function, we show that the limit of infinitely many hidden units is well defined. As with RBM, approximate maximum likelihood training can be performed, resulting in an algorithm that naturally and adaptively adds trained hidden units during learning. We empirically study the behavior of this infinite RBM, showing that its performance is competitive to that of the RBM, while not requiring the tuning of a hidden layer size.

Topic modeling based on latent Dirichlet allocation (LDA) has been a framework of choice to deal with multimodal data, such as in image annotation tasks. Another popular approach to model the multimodal data is through deep neural networks, such as the deep Boltzmann machine (DBM). Recently, a new type of topic model called the Document Neural Autoregressive Distribution Estimator (DocNADE) was proposed and demonstrated state-of-the-art performance for text document modeling. In this work, we show how to successfully apply and extend this model to multimodal data, such as simultaneous image classification and annotation. First, we propose SupDocNADE, a supervised extension of DocNADE, that increases the discriminative power of the learned hidden topic features and show how to employ it to learn a joint representation from image visual words, annotation words and class label information. We test our model on the LabelMe and UIUC-Sports data sets and show that it compares favorably to other topic models. Second, we propose a deep extension of our model and provide an efficient way of training the deep model. Experimental results show that our deep model outperforms its shallow version and reaches state-of-the-art performance on the Multimedia Information Retrieval (MIR) Flickr data set.

Recent progress in using recurrent neural networks (RNNs) for image description has motivated the exploration of their application for video description. However, while images are static, working with videos requires modeling their dynamic temporal structure and then properly integrating that information into a natural language description model. In this context, we propose an approach that successfully takes into account both the local and global temporal structure of videos to produce descriptions. First, our approach incorporates a spatial temporal 3-D convolutional neural network (3-D CNN) representation of the short temporal dynamics. The 3-D CNN representation is trained on video action recognition tasks, so as to produce a representation that is tuned to human motion and behavior. Second we propose a temporal attention mechanism that allows to go beyond local temporal modeling and learns to automatically select the most relevant temporal segments given the text-generating RNN. Our approach exceeds the current state-of-art for both BLEU and METEOR metrics on the Youtube2Text dataset. We also present results on a new, larger and more challenging dataset of paired video and natural language descriptions.

There has been a lot of recent interest in designing neural network models to estimate a distribution from a set of examples. We introduce a simple modification for autoencoder neural networks that yields powerful generative models. Our method masks the autoencoder’s parameters to respect autoregressive constraints: each input is reconstructed only from previous inputs in a given ordering. Constrained this way, the autoencoder outputs can be interpreted as a set of conditional probabilities, and their product, the full joint probability. We can also train a single network that can decompose the joint probability in multiple different orderings. Our simple framework can be applied to multiple architectures, including deep ones. Vectorized implementations, such as on GPUs, are simple and fast. Experiments demonstrate that this approach is competitive with state-of-the-art tractable distribution estimators. At test time, the method is significantly faster and scales better than other autoregressive estimators.

The twenty papers in this special section aim at providing a forum to present recent advancements in deep learning research that directly concerns the multimedia community. Specifically, deep learning has successfully designed algorithms that can build deep nonlinear representations to mimic how the brain perceives and understands multimodal information, ranging from low-level signals like images and audios, to high-level semantic data like natural language. For multimedia research, it is especially important to develop deep networks to capture the dependencies between different genres of data, building joint deep representation for diverse modalities.

Topic modeling based on latent Dirichlet allocation (LDA) has been a framework of choice to deal with multimodal data, such as in image annotation tasks. Recently, a new type of topic model called the Document Neural Autoregressive Distribution Estimator (DocNADE) was proposed and demonstrated state-of-the-art performance for text document modeling. In this work, we show how to successfully apply and extend this model to multimodal data, such as simultaneous image classification and annotation. Specifically, we propose SupDocNADE, a supervised extension of DocNADE, that increases the discriminative power of the hidden topic features by incorporating label information into the training objective of the model and show how to employ SupDocNADE to learn a joint representation from image visual words, annotation words and class label information. We also describe how to leverage information about the spatial position of the visual words for SupDocNADE to achieve better performance in a simple, yet effective manner. We test our model on the LabelMe and UIUC-Sports datasets and show that it compares favorably to other topic models such as the supervised variant of LDA and a Spatial Matching Pyramid (SPM) approach.

We propose a method for producing ensembles of predictors based on holdout estimations of their generalization performances. This approach uses a prior directly on the performance of predictors taken from a finite set of candidates and attempts to infer which one is best. Using Bayesian inference, we can thus obtain a posterior that represents our uncertainty about that choice and construct a weighted ensemble of predictors accordingly. This approach has the advantage of not requiring that the predictors be probabilistic themselves, can deal with arbitrary measures of performance and does not assume that the data was actually generated from any of the predictors in the ensemble. Since the problem of finding the best (as opposed to the true) predictor among a class is known as agnostic PAC-learning, we refer to our method as agnostic Bayesian learning. We also propose a method to address the case where the performance estimate is obtained from k-fold cross validation. While being efficient and easily adjustable to any loss function, our experiments confirm that the agnostic Bayes approach is state of the art compared to common baselines such as model selection based on k-fold cross-validation or a linear combination of predictor outputs.

The Neural Autoregressive Distribution Estimator (NADE) and its real-valued version RNADE are competitive density models of multidimensional data across a variety of domains. These models use a fixed, arbitrary ordering of the data dimensions. One can easily condition on variables at the beginning of the ordering, and marginalize out variables at the end of the ordering, however other inference tasks require approximate inference. In this work we introduce an efficient procedure to simultaneously train a NADE model for each possible ordering of the variables, by sharing parameters across all these models. We can thus use the most convenient model for each inference task at hand, and ensembles of such models with different orderings are immediately available. Moreover, unlike the original NADE, our training procedure scales to deep models. Empirically, ensembles of Deep NADE models obtain state of the art density estimation performance.

We consider the problem of brain tumor segmentation from magnetic resonance (MR) images. This task is most frequently tackled using machine learning methods that generalize across brains, by learning from training brain images in order to generalize to novel test brains. However this approach faces many obstacles that threaten its performance, such as the ability to properly perform multi-brain registration or brain-atlas alignment, or to extract appropriate high-dimensional features that support good generalization. These operations are both nontrivial and time-consuming, limiting the practicality of these approaches in a clinical context. In this paper, we propose to side step these issues by approaching the problem as one of within brain generalization. Specifically, we propose a semi-automatic method that segments a given brain by training and generalizing within that brain only, based on some minimum user interaction. We investigate how k nearest neighbors (kNN), arguably the simplest machine learning method available, combined with the simplest feature vector possible (raw MR signal + (x,y,z) position) can be combined into a method that is both simple, accurate and fast. Results obtained on the online BRATS dataset reveal that our method is fast and second best in terms of the complete and core test set tumor segmentation.

We introduce a novel graphical model, the collaborative score topic model (CSTM), for personal recommendations of textual documents. CSTM's chief novelty lies in its learned model of individual libraries, or sets of documents, associated with each user. Overall, CSTM is a joint directed probabilistic model of user-item scores (ratings), and the textual side information in the user libraries and the items. Creating a generative description of scores and the text allows CSTM to perform well in a wide variety of data regimes, smoothly combining the side information with observed ratings as the number of ratings available for a given user ranges from none to many. Experiments on real-world datasets demonstrate CSTM's performance. We further demonstrate its utility in an application for personal recommendations of posters which we deployed at the NIPS 2013 conference.

We introduce RNADE, a new model for joint density estimation of real-valued vectors. Our model calculates the density of a datapoint as the product of one-dimensional conditionals modeled using mixture density networks with shared parameters. RNADE learns a distributed representation of the data, while having a tractable expression for the calculation of densities. A tractable likelihood allows direct comparison with other methods and training by standard gradient-based optimizers. We compare the performance of RNADE on several datasets of heterogeneous and perceptual data, finding it outperforms mixture models in all but one case.

There has been a resurgence of research in the design of deep architecture models and learning algorithms, i.e., methods that rely on the extraction of a multilayer representation of the data. Often referred to as deep learning, this topic of research has been building on and contributing to many different research topics, such as neural networks, graphical models, feature learning, unsupervised learning, optimization, pattern recognition, and signal processing. Deep learning is also motivated and inspired by neuroscience and has had a tremendous impact on various applications such as computer vision, speech recognition, and natural language processing. The clearly multidisciplinary nature of deep learning led to a call for papers for a special issue dedicated to learning deep architectures.

We present a general treatment of the problem of aggregating preferences from several experts into a consensus ranking, in the context where information about a target ranking is available. Specifically, we describe how such problems can be converted into a standard learning-to-rank one on which existing learning solutions can be invoked. This transformation allows us to optimize the aggregating function for any target IR metric, such as Normalized Discounted Cumulative Gain, or Expected Reciprocal Rank. When applied to crowdsourcing and meta-search benchmarks, our new algorithm improves on state-of-the-art preference aggregation methods.

We describe a new model for learning meaningful representations of text documents from an unlabeled collection of documents. This model is inspired by the recently proposed Replicated Softmax, an undirected graphical model of word counts that was shown to learn a better generative model and more meaningful document representations. Specifically, we take inspiration from the conditional mean-field recursive equations of the Replicated Softmax in order to define a neural network architecture that estimates the probability of observing a new word in a given document given the previously observed words. This paradigm also allows us to replace the expensive softmax distribution over words with a hierarchical distribution over paths in a binary tree of words. The end result is a model whose training complexity scales logarithmically with the vocabulary size instead of linearly as in the Replicated Softmax. Our experiments show that our model is competitive both as a generative model of documents and as a document representation learning algorithm.

The use of machine learning algorithms frequently involves careful tuning of learning parameters and model hyperparameters. Unfortunately, this tuning is often a “black art” requiring expert experience, rules of thumb, or sometimes brute-force search. There is therefore great appeal for automatic approaches that can optimize the performance of any given learning algorithm to the problem at hand. In this work, we consider this problem through the framework of Bayesian optimization, in which a learning algorithm’s generalization performance is modeled as a sample from a Gaussian process (GP). We show that certain choices for the nature of the GP, such as the type of kernel and the treatment of its hyperparameters, can play a crucial role in obtaining a good optimizer that can achieve expert-level performance. We describe new algorithms that take into account the variable cost (duration) of learning algorithm experiments and that can leverage the presence of multiple cores for parallel experimentation. We show that these proposed algorithms improve on previous automatic procedures and can reach or surpass human expert-level optimization for many algorithms including Latent Dirichlet Allocation, Structured SVMs and convolutional neural networks.

We compare the recently proposed Discriminative Restricted Boltzmann Machine (DRBM) to the classical Support Vector Machine (SVM) on a challenging classification task consisting in identifying weapon classes from audio signals. The three weapon classes considered in this work (mortar, rocket, and rocket‐propelled grenade), are difficult to reliably classify with standard techniques because they tend to have similar acoustic signatures. In addition, specificities of the data available in this study make it challenging to rigorously compare classifiers, and we address methodological issues arising from this situation. Experiments show good classification accuracy that could make these techniques suitable for fielding on autonomous devices. DRBMs appear to yield better accuracy than SVMs, and are less sensitive to the choice of signal preprocessing and model hyperparameters. This last property is especially appealing in such a task where the lack of data makes model validation difficult.

We discuss an attentional model for simultaneous object tracking and recognition that is driven by gaze data. Motivated by theories of perception, the model consists of two interacting pathways, identity and control, intended to mirror the what and where pathways in neuroscience models. The identity pathway models object appearance and performs classification using deep (factored)-restricted Boltzmann machines. At each point in time, the observations consist of foveated images, with decaying resolution toward the periphery of the gaze. The control pathway models the location, orientation, scale, and speed of the attended object. The posterior distribution of these states is estimated with particle filtering. Deeper in the control pathway, we encounter an attentional mechanism that learns to select gazes so as to minimize tracking uncertainty. Unlike in our previous work, we introduce gaze selection strategies that operate in the presence of partial information and on a continuous action space. We show that a straightforward extension of the existing approach to the partial information setting results in poor performance, and we propose an alternative method based on modeling the reward surface as a gaussian process. This approach gives good performance in the presence of partial information and allows us to expand the action space from a small, discrete set of fixation points to a continuous domain.

We describe a model based on a Boltzmann machine with third-order connections that can learn how to accumulate information about a shape over several fixations. The model uses a retina that only has enough high resolution pixels to cover a small area of the image, so it must decide on a sequence of fixations and it must combine the glimpse" at each fixation with the location of the fixation before integrating the information with information from other glimpses of the same object. We evaluate this model on a synthetic dataset and two image classification datasets, showing that it can perform at least as well as a model trained on whole images."

We investigate the problem of estimating the density function of multivariate binary data. In particular, we focus on models for which computing the estimated probability of any data point is tractable. In such a setting, previous work has mostly concentrated on mixture modeling approaches. We argue that for the problem of tractable density estimation, the restricted Boltzmann machine (RBM) provides a competitive framework for multivariate binary density modeling. With this in mind, we also generalize the RBM framework and present the restricted Boltzmann forest (RBForest), which replaces the binary variables in the hidden layer of RBMs with groups of tree-structured binary variables. This extension allows us to obtain models that have more modeling capacity but remain tractable. In experiments on several data sets, we demonstrate the competitiveness of this approach and study some of its properties.

Recently, many applications for Restricted Boltzmann Machines (RBMs) have been developed for a large variety of learning problems. However, RBMs are usually used as feature extractors for another learning algorithm or to provide a good initialization for deep feed-forward neural network classifiers, and are not considered as a standalone solution to classification problems. In this paper, we argue that RBMs provide a self-contained framework for deriving competitive non-linear classifiers. We present an evaluation of different learning algorithms for RBMs which aim at introducing a discriminative component to RBM training and improve their performance as classifiers. This approach is simple in that RBMs are used directly to build a classifier, rather than as a stepping stone. Finally, we demonstrate how discriminative RBMs can also be successfully employed in a semi-supervised setting.

Previous work has shown that the difficulties in learning deep generative or discriminative models can be overcome by an initial unsupervised learning step that maps inputs to useful intermediate representations. We introduce and motivate a new training principle for unsupervised learning of a representation based on the idea of making the learned representations robust to partial corruption of the input pattern. This approach can be used to train autoencoders, and these denoising autoencoders can be stacked to initialize deep architectures. The algorithm can be motivated from a manifold learning and information theoretic perspective or from a generative model perspective. Comparative experiments clearly show the surprising advantage of corrupting the input of autoencoders on a pattern classification benchmark suite. 1.

Recently, several learning algorithms relying on models with deep architectures have been proposed. Though they have demonstrated impressive performance, to date, they have only been evaluated on relatively simple problems such as digit recognition in a controlled environment, for which many machine learning algorithms already report reasonable results. Here, we present a series of experiments which indicate that these models show promise in solving harder learning problems that exhibit many factors of variation. These models are compared with well-established algorithms such as Support Vector Machines and single hidden-layer feed-forward neural networks.

Recent analyses (Bengio, Delalleau, & Le Roux, 2006; Bengio & Le Cun, 2007) of modern nonparametric machine learning algorithms that are kernel machines, such as Support Vector Machines (SVMs), graph-based manifold and semi-supervised learning algorithms suggest fundamental limitations of some learning algorithms. The problem is clear in kernel-based approaches when the kernel is "local" (e.g., the Gaussian kernel), i.e., K (x, y ) converges to a constant when ||x - y || increases. These analyses point to the difficulty of learning "highly-varying functions", i.e., functions that have a large number of "variations" in the domain of interest, e.g., they would require a large number of pieces to be well represented by a piecewise-linear approximation. Since the number of pieces can be made to grow exponentially with the number of factors of variations in the input, this is connected with the well-known curse of dimensionality for classical non-parametric learning algorithms (for regression, classification and density estimation). If the shapes of all these pieces are unrelated, one needs enough examples for each piece in order to generalize properly. However, if these shapes are related and can be predicted from each other, "non-local" learning algorithms have the potential to generalize to pieces not covered by the training set. Such ability would seem necessary for learning in complex domains such as Artificial Intelligence tasks (e.g., related to vision, language, speech, robotics). Kernel machines (not only those with a local kernel) have a shallow architecture, i.e., only two levels of data-dependent computational elements. This is also true of feedforward neural networks with a single hidden layer (which can become SVMs when the number of hidden units becomes large (Bengio, Le Roux, Vincent, Delalleau, & Marcotte, 2006)). A serious problem with shallow architectures is that they can be very inefficient in terms of the number of computational units (e.g., bases, hidden units), and thus in terms of required examples (Bengio & Le Cun, 2007). One way to represent a highly-varying function compactly (with few parameters) is through the composition of many non-linearities, i.e., with a deep architecture. For example, the parity function with d inputs requires O(2d ) examples and parameters to be represented by a Gaussian SVM (Bengio et al., 2006), O(d2 ) parameters for a one-hidden-layer neural network, O(d) parameters and units for a multi-layer network with O(log2 d) layers, and O(1) parameters with a recurrent neural network. More generally, boolean functions (such as the function that computes the multiplication of two numbers from their d-bit representation) expressible by O(log d) layers of combinatorial logic with O(d) elements in each layer may require O(2d ) elements when expressed with only 2 layers (Utgoff & Stracuzzi, 2002; Bengio & Le Cun, 2007). When the representation of a concept requires an exponential number of elements, e.g., with a shallow circuit, the number of training examples required to learn the concept may also be impractical. Formal analyses of the computational complexity of shallow circuits can be found in (Hastad, 1987) or (Allender, 1996). They point in the same direction: shallow circuits are much less expressive than deep ones. However, until recently, it was believed too difficult to train deep multi-layer neural networks. Empirically, deep networks were generally found to be not better, and often worse, than neural networks with one or two hidden layers (Tesauro, 1992). As this is a negative result, it has not been much reported in the machine learning literature. A reasonable explanation is that gradient-based optimization starting from random initialization may get stuck near poor solutions. An approach that has been explored with some success in the past is based on constructively adding layers. This was previously done using a supervised criterion at each stage (Fahlman & Lebiere, 1990; Lengelle & Denoeux, 1996). Hinton, Osindero, and Teh (2006) recently introduced a greedy layer-wise unsupervised learning algorithm for Deep Belief Networks (DBN), a generative model with many layers of hidden causal variables. The training strategy for such networks may hold great promise as a principle to help address the problem of training deep networks. Upper layers of a DBN are supposed to represent more "abstract" concepts that explain the input observation x, whereas lower layers extract "low-level features" from x. They learn simpler concepts first, and build on them to learn more abstract concepts. This strategy, studied in detail here, has not yet been much exploited in machine learning. We hypothesize that three aspects of this strategy are particularly important: first, pre-training one layer at a time in a greedy way; second, using unsupervised learning at each layer in order to preserve information from the input; and finally, fine-tuning the whole network with respect to the ultimate criterion of interest. We first extend DBNs and their component layers, Restricted Boltzmann Machines (RBM), so that they can more naturally handle continuous values in input. Second, we perform experiments to better understand the advantage brought by the greedy layer-wise unsupervised learning. The basic question to answer is whether or not this approach helps to solve a difficult optimization problem. In DBNs, RBMs are used as building blocks, but applying this same strategy using auto-encoders yielded similar results. Finally, we discuss a problem that occurs with the layer-wise greedy unsupervised procedure when the input distribution is not revealing enough of the conditional distribution of the target variable given the input variable. We evaluate a simple and successful solution to this problem. 2 Deep Belief Nets Let x be the input, and gi the hidden variables at layer i, with joint distribution P (x, g1 , g2 , . . . , g ) = P (x|g1 )P (g1 |g2 ) P (g

We claim and present arguments to the effect that a large class of manifold learning algorithms that are essentially local and can be framed as kernel learning algorithms will suffer from the curse of dimensionality, at the dimension of the true underlying manifold. This observation invites an exploration of nonlocal manifold learning algorithms that attempt to discover shared structure in the tangent planes at different positions. A training criterion for such an algorithm is proposed, and experiments estimating a tangent plane prediction function are presented, showing its advantages with respect to local manifold learning algorithms: it is able to generalize very far from training data (on learning handwritten character image rotations), where local nonparametric methods fail.

To escape from the curse of dimensionality, we claim that one can learn non-local functions, in the sense that the value and shape of the learned function at x must be inferred using examples that may be far from x. With this objective, we present a non-local non-parametric density esti- mator. It builds upon previously proposed Gaussian mixture models with regularized covariance matrices to take into account the local shape of the manifold. It also builds upon recent work on non-local estimators of the tangent plane of a manifold, which are able to generalize in places with little training data, unlike traditional, local, non-parametric models.