Pseudo labeling (PL) is a wide-applied strategy to enlarge the labeled dataset by self-annotating the potential samples during the training process. Several works have shown that it can improve the graph learning model performance in general. However, we notice that the incorrect labels can be fatal to the graph training process. Inappropriate PL may result in the performance degrading, especially on graph data where the noise can propagate. Surprisingly, the corresponding error is seldom theoretically analyzed in the literature. In this paper, we aim to give deep insights of PL on graph learning models. We first present the error analysis of PL strategy by showing that the error is bounded by the confidence of PL threshold and consistency of multi-view prediction. Then, we theoretically illustrate the effect of PL on convergence property. Based on the analysis, we propose a cautious pseudo labeling methodology in which we pseudo label the samples with highest confidence and multi-view consistency. Finally, extensive experiments demonstrate that the proposed strategy improves graph learning process and outperforms other PL strategies on link prediction and node classification tasks.

The discrete logarithm problem is a fundamental challenge in number theory with significant implications for cryptographic protocols. In this paper, we investigate the limitations of gradient-based methods for learning the parity bit of the discrete logarithm in finite cyclic groups of prime order. Our main result, supported by theoretical analysis and empirical verification, reveals the concentration of the gradient of the loss function around a fixed point, independent of the logarithm's base used. This concentration property leads to a restricted ability to learn the parity bit efficiently using gradient-based methods, irrespective of the complexity of the network architecture being trained. Our proof relies on Boas-Bellman inequality in inner product spaces and it involves establishing approximate orthogonality of discrete logarithm's parity bit functions through the spectral norm of certain matrices. Empirical experiments using a neural network-based approach further verify the limitations of gradient-based learning, demonstrating the decreasing success rate in predicting the parity bit as the group order increases.

In recent years, machine learning has found gainful applications in diverse domains, but it still has a heavy dependence on expensive labeled data: Advances in cheap computing and storage have made it easier to store and process large amounts of unlabeled data, but the labeling often needs to be done by humans or using costly tools. Therefore, there is a need to develop general domain-independent methods to learn models effectively from a large amount of unlabeled data at the disposal, along with a minimal amount of labeled data: this is the framework of semi-supervised learning. Active learning aims explicitly to detect the observations to be labeled to optimize the learning process and efficiently reduce the labeling cost. The primary assumption behind active learning is that machine learning algorithms could reach a higher level of performance while using a smaller number of training labels if they were allowed to choose the training dataset (Settles, 2009). The most common active learning approaches are pool-based methods (Lewis and Catlett, 1994) based on a set of unlabeled observations. First, some points are labeled to train a classification model, and then, at each iteration, we choose unlabeled examples to query based on the predictions of the current model and a predefined priority score. These approaches show their limitations in low-budget regime scenarios because they need a sufficient budget to learn a weak model (Pourahmadi et al., 2021). 1

This study proposes a novel memory-efficient recurrent neural network (RNN) architecture specified to solve the object localization problem. This problem is to recover the object states along with its movement in a noisy environment. We take the idea of the classical particle filter and combine it with GRU RNN architecture. The key feature of the resulting memory-efficient particle filter RNN model (mePFRNN) is that it requires the same number of parameters to process environments of different sizes. Thus, the proposed mePFRNN architecture consumes less memory to store parameters compared to the previously proposed PFRNN model. To demonstrate the performance of our model, we test it on symmetric and noisy environments that are incredibly challenging for filtering algorithms. In our experiments, the mePFRNN model provides more precise localization than the considered competitors and requires fewer trained parameters.

Training AI models that generalize across tasks and domains has long been among the open problems driving AI research. The emergence of Foundation Models made it easier to obtain expert models for a given task, but the heterogeneity of data that may be encountered at test time often means that any single expert is insufficient. We consider the Fusion of Experts (FoE) problem of fusing outputs of expert models with complementary knowledge of the data distribution and formulate it as an instance of supervised learning. Our method is applicable to both discriminative and generative tasks and leads to significant performance improvements in image and text classification, text summarization, multiple-choice QA, and automatic evaluation of generated text. We also extend our method to the "frugal" setting where it is desired to reduce the number of expert model evaluations at test time.

The Internet of Federated Things (IoFT) represents a network of interconnected systems with federated learning as the backbone, facilitating collaborative knowledge acquisition while ensuring data privacy for individual systems. The wide adoption of IoFT, however, is hindered by security concerns, particularly the susceptibility of federated learning networks to adversarial attacks. In this paper, we propose an effective non-parametric approach FedRR, which leverages the low-rank features of the transmitted parameter updates generated by federated learning to address the adversarial attack problem. Besides, our proposed method is capable of accurately detecting adversarial clients and controlling the false alarm rate under the scenario with no attack occurring. Experiments based on digit recognition using the MNIST datasets validated the advantages of our approach.

Catastrophic forgetting remains a significant challenge to continual learning for decades. While recent works have proposed effective methods to mitigate this problem, they mainly focus on the algorithmic side. Meanwhile, we do not fully understand what architectural properties of neural networks lead to catastrophic forgetting. This study aims to fill this gap by studying the role of activation functions in the training dynamics of neural networks and their impact on catastrophic forgetting. Our study reveals that, besides sparse representations, the gradient sparsity of activation functions also plays an important role in reducing forgetting. Based on this insight, we propose a new class of activation functions, elephant activation functions, that can generate both sparse representations and sparse gradients. We show that by simply replacing classical activation functions with elephant activation functions, we can significantly improve the resilience of neural networks to catastrophic forgetting. Our method has broad applicability and benefits for continual learning in regression, class incremental learning, and reinforcement learning tasks. Specifically, we achieves excellent performance on Split MNIST dataset in just one single pass, without using replay buffer, task boundary information, or pre-training. One of the biggest challenges to achieving continual learning is the decades-old issue of catastrophic forgetting (French 1999). Catastrophic forgetting stands for the phenomenon that artificial neural networks tend to forget prior knowledge drastically when learned with stochastic gradient descent algorithms on non-independent and identically distributed (non-iid) data.

Fleets of robots ingest massive amounts of streaming data generated by interacting with their environments, far more than those that can be stored or transmitted with ease. At the same time, we hope that teams of robots can co-acquire diverse skills through their experiences in varied settings. How can we enable such fleet-level learning without having to transmit or centralize fleet-scale data? In this paper, we investigate distributed learning of policies as a potential solution. To efficiently merge policies in the distributed setting, we propose fleet-merge, an instantiation of distributed learning that accounts for the symmetries that can arise in learning policies that are parameterized by recurrent neural networks. We show that fleet-merge consolidates the behavior of policies trained on 50 tasks in the Meta-World environment, with the merged policy achieving good performance on nearly all training tasks at test time. Moreover, we introduce a novel robotic tool-use benchmark, fleet-tools, for fleet policy learning in compositional and contact-rich robot manipulation tasks, which might be of broader interest, and validate the efficacy of fleet-merge on the benchmark.

Federated learning is a technique that enables the use of distributed datasets for machine learning purposes without requiring data to be pooled, thereby better preserving privacy and ownership of the data. While supervised FL research has grown substantially over the last years, unsupervised FL methods remain scarce. This work introduces an algorithm which implements K-means clustering in a federated manner, addressing the challenges of varying number of clusters between centers, as well as convergence on less separable datasets.

Despite the recent advances in the field of computational Schrodinger Bridges (SB), most existing SB solvers are still heavy-weighted and require complex optimization of several neural networks. It turns out that there is no principal solver which plays the role of simple-yet-effective baseline for SB just like, e.g., $k$-means method in clustering, logistic regression in classification or Sinkhorn algorithm in discrete optimal transport. We address this issue and propose a novel fast and simple SB solver. Our development is a smart combination of two ideas which recently appeared in the field: (a) parameterization of the Schrodinger potentials with sum-exp quadratic functions and (b) viewing the log-Schrodinger potentials as the energy functions. We show that combined together these ideas yield a lightweight, simulation-free and theoretically justified SB solver with a simple straightforward optimization objective. As a result, it allows solving SB in moderate dimensions in a matter of minutes on CPU without a painful hyperparameter selection. Our light solver resembles the Gaussian mixture model which is widely used for density estimation. Inspired by this similarity, we also prove an important theoretical result showing that our light solver is a universal approximator of SBs. The code for the LightSB solver can be found at https://github.com/ngushchin/LightSB