Adversarial Training for High-Stakes Reliability

In the future, powerful AI systems may be deployed in high-stakes settings, where a single failure could be catastrophic. One technique for improving AI safety in high-stakes settings is adversarial training, which uses an adversary to generate examples to train on in order to achieve better worst-case performance. In this work, we used a language generation task as a testbed for achieving high reliability through adversarial training. We created a series of adversarial training techniques -- including a tool that assists human adversaries -- to find and eliminate failures in a classifier that filters text completions suggested by a generator. In our simple "avoid injuries" task, we determined that we can set very conservative classifier thresholds without significantly impacting the quality of the filtered outputs. With our chosen thresholds, filtering with our baseline classifier decreases the rate of unsafe completions from about 2.4% to 0.003% on in-distribution data, which is near the limit of our ability to measure. We found that adversarial training significantly increased robustness to the adversarial attacks that we trained on, without affecting in-distribution performance. We hope to see further work in the high-stakes reliability setting, including more powerful tools for enhancing human adversaries and better ways to measure high levels of reliability, until we can confidently rule out the possibility of catastrophic deployment-time failures of powerful models.

Learning Disentangled Behaviour Patterns for Wearable-based Human Activity Recognition

In wearable-based human activity recognition (HAR) research, one of the major challenges is the large intra-class variability problem. The collected activity signal is often, if not always, coupled with noises or bias caused by personal, environmental, or other factors, making it difficult to learn effective features for HAR tasks, especially when with inadequate data. To address this issue, in this work, we proposed a Behaviour Pattern Disentanglement (BPD) framework, which can disentangle the behavior patterns from the irrelevant noises such as personal styles or environmental noises, etc. Based on a disentanglement network, we designed several loss functions and used an adversarial training strategy for optimization, which can disentangle activity signals from the irrelevant noises with the least dependency (between them) in the feature space. Our BPD framework is flexible, and it can be used on top of existing deep learning (DL) approaches for feature refinement. Extensive experiments were conducted on four public HAR datasets, and the promising results of our proposed BPD scheme suggest its flexibility and effectiveness. This is an open-source project, and the code can be found at http://github.com/Jie-su/BPD

An Improved Analysis of Gradient Tracking for Decentralized Machine Learning

We consider decentralized machine learning over a network where the training data is distributed across $n$ agents, each of which can compute stochastic model updates on their local data. The agent's common goal is to find a model that minimizes the average of all local loss functions. While gradient tracking (GT) algorithms can overcome a key challenge, namely accounting for differences between workers' local data distributions, the known convergence rates for GT algorithms are not optimal with respect to their dependence on the mixing parameter $p$ (related to the spectral gap of the connectivity matrix). We provide a tighter analysis of the GT method in the stochastic strongly convex, convex and non-convex settings. We improve the dependency on $p$ from $\mathcal{O}(p^{-2})$ to $\mathcal{O}(p^{-1}c^{-1})$ in the noiseless case and from $\mathcal{O}(p^{-3/2})$ to $\mathcal{O}(p^{-1/2}c^{-1})$ in the general stochastic case, where $c \geq p$ is related to the negative eigenvalues of the connectivity matrix (and is a constant in most practical applications). This improvement was possible due to a new proof technique which could be of independent interest.

Towards Federated Learning on Time-Evolving Heterogeneous Data

Federated Learning (FL) is an emerging learning paradigm that preserves privacy by ensuring client data locality on edge devices. The optimization of FL is challenging in practice due to the diversity and heterogeneity of the learning system. Despite recent research efforts on improving the optimization of heterogeneous data, the impact of time-evolving heterogeneous data in real-world scenarios, such as changing client data or intermittent clients joining or leaving during training, has not been well studied. In this work, we propose Continual Federated Learning (CFL), a flexible framework, to capture the time-evolving heterogeneity of FL. CFL covers complex and realistic scenarios -- which are challenging to evaluate in previous FL formulations -- by extracting the information of past local datasets and approximating the local objective functions. Theoretically, we demonstrate that CFL methods achieve a faster convergence rate than \fedavg in time-evolving scenarios, with the benefit being dependent on approximation quality. In a series of experiments, we show that the numerical findings match the convergence analysis, and CFL methods significantly outperform the other SOTA FL baselines.

Understanding Memorization from the Perspective of Optimization via Efficient Influence Estimation

Over-parameterized deep neural networks are able to achieve excellent training accuracy while maintaining a small generalization error. It has also been found that they are able to fit arbitrary labels, and this behaviour is referred to as the phenomenon of memorization. In this work, we study the phenomenon of memorization with turn-over dropout, an efficient method to estimate influence and memorization, for data with true labels (real data) and data with random labels (random data). Our main findings are: (i) For both real data and random data, the optimization of easy examples (e.g., real data) and difficult examples (e.g., random data) are conducted by the network simultaneously, with easy ones at a higher speed; (ii) For real data, a correct difficult example in the training dataset is more informative than an easy one. By showing the existence of memorization on random data and real data, we highlight the consistency between them regarding optimization and we emphasize the implication of memorization during optimization.

Learning by Active Forgetting for Neural Networks

Remembering and forgetting mechanisms are two sides of the same coin in a human learning-memory system. Inspired by human brain memory mechanisms, modern machine learning systems have been working to endow machine with lifelong learning capability through better remembering while pushing the forgetting as the antagonist to overcome. Nevertheless, this idea might only see the half picture. Up until very recently, increasing researchers argue that a brain is born to forget, i.e., forgetting is a natural and active process for abstract, rich, and flexible representations. This paper presents a learning model by active forgetting mechanism with artificial neural networks. The active forgetting mechanism (AFM) is introduced to a neural network via a "plug-and-play" forgetting layer (P\&PF), consisting of groups of inhibitory neurons with Internal Regulation Strategy (IRS) to adjust the extinction rate of themselves via lateral inhibition mechanism and External Regulation Strategy (ERS) to adjust the extinction rate of excitatory neurons via inhibition mechanism. Experimental studies have shown that the P\&PF offers surprising benefits: self-adaptive structure, strong generalization, long-term learning and memory, and robustness to data and parameter perturbation. This work sheds light on the importance of forgetting in the learning process and offers new perspectives to understand the underlying mechanisms of neural networks.

RelaySum for Decentralized Deep Learning on Heterogeneous Data

In decentralized machine learning, workers compute model updates on their local data. Because the workers only communicate with few neighbors without central coordination, these updates propagate progressively over the network. This paradigm enables distributed training on networks without all-to-all connectivity, helping to protect data privacy as well as to reduce the communication cost of distributed training in data centers. A key challenge, primarily in decentralized deep learning, remains the handling of differences between the workers' local data distributions. To tackle this challenge, we introduce the RelaySum mechanism for information propagation in decentralized learning. RelaySum uses spanning trees to distribute information exactly uniformly across all workers with finite delays depending on the distance between nodes. In contrast, the typical gossip averaging mechanism only distributes data uniformly asymptotically while using the same communication volume per step as RelaySum. We prove that RelaySGD, based on this mechanism, is independent of data heterogeneity and scales to many workers, enabling highly accurate decentralized deep learning on heterogeneous data. Our code is available at http://github.com/epfml/relaysgd.

Nash Convergence of Mean-Based Learning Algorithms in First Price Auctions

We consider repeated first price auctions where each bidder, having a deterministic type, learns to bid using a mean-based learning algorithm. We completely characterize the Nash convergence property of the bidding dynamics in two senses: (1) time-average: the fraction of rounds where bidders play a Nash equilibrium approaches to 1 in the limit; (2) last-iterate: the mixed strategy profile of bidders approaches to a Nash equilibrium in the limit. Specifically, the results depend on the number of bidders with the highest value: - If the number is at least three, the bidding dynamics almost surely converges to a Nash equilibrium of the auction, both in time-average and in last-iterate. - If the number is two, the bidding dynamics almost surely converges to a Nash equilibrium in time-average but not necessarily in last-iterate. - If the number is one, the bidding dynamics may not converge to a Nash equilibrium in time-average nor in last-iterate. Our discovery opens up new possibilities in the study of convergence dynamics of learning algorithms.

Representation Memorization for Fast Learning New Knowledge without Forgetting

The ability to quickly learn new knowledge (e.g. new classes or data distributions) is a big step towards human-level intelligence. In this paper, we consider scenarios that require learning new classes or data distributions quickly and incrementally over time, as it often occurs in real-world dynamic environments. We propose "Memory-based Hebbian Parameter Adaptation" (Hebb) to tackle the two major challenges (i.e., catastrophic forgetting and sample efficiency) towards this goal in a unified framework. To mitigate catastrophic forgetting, Hebb augments a regular neural classifier with a continuously updated memory module to store representations of previous data. To improve sample efficiency, we propose a parameter adaptation method based on the well-known Hebbian theory, which directly "wires" the output network's parameters with similar representations retrieved from the memory. We empirically verify the superior performance of Hebb through extensive experiments on a wide range of learning tasks (image classification, language model) and learning scenarios (continual, incremental, online). We demonstrate that Hebb effectively mitigates catastrophic forgetting, and it indeed learns new knowledge better and faster than the current state-of-the-art.