Aligning diffusion model outputs with downstream objectives is essential for improving task-specific performance. Broadly, inference-time training-free approaches for aligning diffusion models can be categorized into two main strategies: sampling-based methods, which explore multiple candidate outputs and select those with higher reward signals, and gradient-guided methods, which use differentiable reward approximations to directly steer the generation process. In this work, we propose a universal algorithm, UniCoDe, which brings together the strengths of sampling and gradient-based guidance into a unified framework. UniCoDe integrates local gradient signals during sampling, thereby addressing the sampling inefficiency inherent in complex reward-based sampling approaches. By cohesively combining these two paradigms, UniCoDe enables more efficient sampling while offering better tradeoffs between reward alignment and divergence from the diffusion unconditional prior. Empirical results demonstrate that UniCoDe remains competitive with state-of-the-art baselines across a range of tasks. The code is available at https://github.com/maurya-goyal10/UniCoDe

Masked Autoencoders (MAEs) are an important di-vide in self-supervised learning (SSL) due to their in-dependence from augmentation techniques for generating positive (and/or negative) pairs as in contrastive frame-works. Their masking and reconstruction strategy also nicely aligns with SSL approaches in natural language pro-cessing. Most MAEs are built upon Transformer-based ar-chitectures where visual features are not regularized as op-posed to their convolutional neural network (CNN) based counterparts, which can potentially hinder their performance. To address this, we introduce MAGMA, a novel batch-wide layer-wise regularization loss applied to rep-resentations of different Transformer layers. We demon-strate that by plugging in the proposed regularization loss, one can significantly improve the performance of MAE-based models. We further demonstrate the impact of the proposed loss on optimizing other generic SSL approaches (such as VICReg and SimCLR), broadening the impact of the proposed approach. Our code base can be found here: https://github.com/adondera/magma

Learning quickly from very few labeled samples is a fundamental attribute that separates machines and humans in the era of deep representation learning. Unsupervised few-shot learning (U-FSL) aspires to bridge this gap by discarding the reliance on annotations at training time. Intrigued by the success of contrastive learning approaches in the realm of U-FSL, we structurally approach their shortcomings in both pretraining and downstream inference stages. We propose a novel Dynamic Clustered mEmory (DyCE) module to promote a highly separable latent representation space for enhancing positive sampling at the pretraining phase and infusing implicit class-level insights into unsupervised contrastive learning. We then tackle the, somehow overlooked yet critical, issue of sample bias at the few-shot inference stage. We propose an iterative Optimal Transport-based distribution Alignment (OpTA) strategy and demonstrate that it efficiently addresses the problem, especially in low-shot scenarios where FSL approaches suffer the most from sample bias. We later on discuss that DyCE and OpTA are two intertwined pieces of a novel end-to-end approach (we coin as BECLR), constructively magnifying each other's impact. We then present a suite of extensive quantitative and qualitative experimentation to corroborate that BECLR sets a new state-of-the-art across ALL existing U-FSL benchmarks (to the best of our knowledge), and significantly outperforms the best of the current baselines (codebase available at https://github.com/stypoumic/BECLR).

Binary Neural Networks (BNNs) are receiving an up-surge of attention for bringing power-hungry deep learning towards edge devices. The traditional wisdom in this space is to employ sign(.) for binarizing feature maps. We argue and illustrate that sign(.) is a uniqueness bottleneck, limiting information propagation throughout the network. To alleviate this, we propose to dispense sign(.), replacing it with a learnable activation binarizer (LAB), allowing the network to learn a fine-grained binarization kernel per layer - as opposed to global thresholding. LAB is a novel universal module that can seamlessly be integrated into existing architectures. To confirm this, we plug it into four seminal BNNs and show a considerable accuracy boost at the cost of tolerable increase in delay and complexity. Finally, we build an end-to-end BNN (coined as LAB-BNN) around LAB, and demonstrate that it achieves competitive performance on par with the state-of-the-art on ImageNet. Our code can be found in our repository: https://github.com/sfalkena/LAB.

Humans have a unique ability to learn new representations from just a handful of examples with little to no supervision. Deep learning models, however, require an abundance of data and supervision to perform at a satisfactory level. Unsupervised few-shot learning (U-FSL) is the pursuit of bridging this gap between machines and humans. Inspired by the capacity of graph neural networks (GNNs) in discovering complex inter-sample relationships, we propose a novel self-attention based message passing contrastive learning approach (coined as SAMP-CLR) for U-FSL pre-training. We also propose an optimal transport (OT) based fine-tuning strategy (we call OpT-Tune) to efficiently induce task awareness into our novel end-to-end unsupervised few-shot classification framework (SAMPTransfer). Our extensive experimental results corroborate the efficacy of SAMPTransferin a variety of downstream few-shot classification scenarios, setting a new state-of-the-art for U-FSL on both miniImageNet and tieredImageNet benchmarks, offering up to 7%+ and 5%+ improvements, respectively. Our further investigations also confirm that SAMPTransferremains on-par with some supervised baselines on miniImageNet and outperforms all existing U-FSL baselines in a challenging cross-domain scenario. Our code can be found in our GitHub repository: https://github.com/ojss/SAMPTransfer/.

Classical federated learning approaches incur significant performance degradation in the presence of non-independent and identically distributed (non-IID) client data. A possible direction to address this issue is forming clusters of clients with roughly IID data. Most solutions following this direction are iterative and relatively slow, also prone to convergence issues in discovering underlying cluster formations. We introduce federated learning with taskonomy (FLT) that generalizes this direction by learning the task relatedness between clients for more efficient federated aggregation of heterogeneous data. In a one-off process, the server provides the clients with a pretrained (and fine-tunable) encoder to compress their data into a latent representation and transmit the signature of their data back to the server. The server then learns the task relatedness among clients via manifold learning and performs a generalization of federated averaging. FLT can flexibly handle a generic client relatedness graph, when there are no explicit clusters of clients, as well as efficiently decompose it into (disjoint) clusters for clustered federated learning. We demonstrate that FLT not only outperforms the existing state-of-the-art baselines in non-IID scenarios but also offers improved fairness across clients. Our codebase can be found at: https://github.com/hjraad/FLT/.

Unsupervised learning is argued to be the dark matter of human intelligence1. To build in this direction, this paper focuses on unsupervised learning from an abundance of unlabeled data followed by few-shot fine-tuning on a downstream classification task. To this aim, we extend a recent study on adopting contrastive learning for self-supervised pre-training by incorporating class-level cognizance through iterative clustering and re-ranking and by expanding the contrastive optimization loss to account for it. To our knowledge, our experimentation both in standard and cross-domain scenarios demonstrate that we set a new state-of-the-art (SoTA) in (5-way, 1 and 5-shot) settings of standard mini-ImageNet benchmark as well as the (5-way, 5 and 20-shot) settings of cross-domain CDFSL benchmark. Our code and experimentation can be found in our GitHub repository: https://github.com/ojss/c3lr.

Semantic segmentation is one of the most fundamental problems in computer vision with significant impact on a wide variety of applications. Adversarial learning is shown to be an effective approach for improving semantic segmentation quality by enforcing higher-level pixel correlations and structural information. However, state-of-the-art semantic segmentation models cannot be easily plugged into an adversarial setting because they are not designed to accommodate convergence and stability issues in adversarial networks. We bridge this gap by building a conditional adversarial network with a state-of-the-art segmentation model (DeepLabv3+) at its core. To battle the stability issues, we introduce a novel lookahead adversarial learning (LoAd) approach with an embedded label map aggregation module. We focus on semantic segmentation models that run fast at inference for near real-time field applications. Through extensive experimentation, we demonstrate that the proposed solution can alleviate divergence issues in an adversarial semantic segmentation setting and results in considerable performance improvements (+5% in some classes) on the baseline for three standard datasets.

The selection of the minimum number of sensors within a network to satisfy a certain estimation performance metric is an interesting problem with a plethora of applications. We explore the sparsity embedded within the problem and propose a relaxed sparsity-aware sensor selection approach which is equivalent to the unrelaxed problem under certain conditions. We also present a reasonably low-complexity and elegant distributed version of the centralized problem with convergence guarantees such that each sensor can decide itself whether it should contribute to the estimation or not. Our simulation results corroborate our claims and illustrate a promising performance for the proposed centralized and distributed algorithms.

The problem of source localization from time-difference-of-arrival (TDOA) measurements is in general a non-convex and complex problem due to its hyperbolic nature. This problem becomes even more complicated for the case of multi-source localization where TDOAs should be assigned to their respective sources. We simplify this problem to an ℓ1-norm minimization by introducing a novel TDOA fingerprinting model for a multi-source scenario. Moreover, we propose an innovative trick to enhance the performance of our proposed fingerprinting model in terms of the number of identifiable sources. An interesting by-product of this enhanced model is that under some conditions we can convert the given underdetermined problem to an overdetermined one and efficiently solve it using classical least squares (LS) approaches. Our simulation results illustrate a good performance for the introduced TDOA fingerprinting.

We extend one of our recently proposed anchorless mobile network localization algorithms (called PEST) to operate in a partially connected network. To this aim, we propose a geometric missing link reconstruction algorithm for noisy scenarios and repeat the proposed algorithm in a local-to-global fashion to reconstruct a complete distance matrix. This reconstructed matrix is then used in the PEST to localize the mobile network. We compare the computational complexity of the new link reconstruction algorithm with existing related algorithms and show that our proposed algorithm has the lowest complexity, and hence, is the best extension of the low complexity PEST. Simulation results further illustrate that the proposed link reconstruction algorithm leads to the lowest reconstruction error as well as the most accurate network localization performance.

In this paper, we analyze the problem of localizing and tracking a mobile node in an underwater environment with an isogradient sound speed profile (SSP). We will show that range-based localization algorithms are not so accurate in such an environment, and they should be replaced by time-based ones. Therefore, we relate the mobile node location to the travel time of a propagating sound wave from (to) an anchor node to (from) the mobile node. After obtaining sufficient time measurements, positioning can be achieved through multilateration. To accomplish this, we utilize the extended Kalman filter (EKF) for multilateration and tracking the mobile node's location in a recursive manner. Through several simulations, we will show that the proposed EKF algorithm performs superb in comparison with algorithms which assume a straight-line wave propagation in an underwater environment.

Two novel cooperative localization algorithms for mobile wireless networks are proposed. To continuously localize the mobile network, given the pairwise distance measurements between different wireless sensor nodes, we propose to use subspace tracking to track the variations in signal eigenvectors and corresponding eigenvalues of the double-centered distance matrix. We compare the computational complexity of the new algorithms with a recently developed algorithm exploiting the extended Kalman filter (EKF) and show that our proposed algorithms are computationally efficient, and hence, appropriate for practical implementations compared to the EKF. Simulation results further illustrate that the proposed algorithms are more accurate when the distance errors are small (low noise scenarios) in comparison with the EKF, while being more robust to the sampling period in high noise scenarios.

We study the problem of optimizing the symbol error probability (SEP) performance of cluster-based cooperative wireless sensor networks (WSNs). It is shown in the literature that an efficient relay selection protocol based on simple geographical information of the nodes to execute the cooperative diversity transmission, can significantly improve the SEP performance at the destination of such networks. Also, a similar analysis on the optimal power allocation for the source and the relay exists in the literature. However, we propose that to achieve the minimum SEP at the destination, a joint optimization of power allocation and relay selection should be accomplished. To this aim, we reformulate the SEP performance at the destination of a simple cluster-based cooperative WSN in the general form and solve the mentioned joint optimization problem efficiently for a relay deployment scenario. Simulation results demonstrate that the proposed joint optimization can effectively improve the SEP performance of the network for both 1-hop and multi-hop scenarios in comparison with the previous disjoint optimal relay deployment scheme.