New methods improve Byzantine robustness in distributed learning.
problem Existing robust aggregation rules fail in realistic scenarios.
method Introducing new robust iterative clipping procedure and worker momentum.
result First provably robust method for standard stochastic optimization.
New bucketing scheme improves Byzantine robustness for heterogeneous data.
problem Byzantine attacks on federated learning with heterogeneous data.
method Bucketing scheme to adapt robust algorithms to non-iid data.
result Bucketing scheme ensures convergence against Byzantine attacks.
New algorithm identifies near-optimal policies in adversarial distributed RL settings.
problem Adversarial agents in distributed RL settings that can collude and report arbitrary data.
method Weighted-Clique algorithm for robust mean estimation from batches, combined with novel distributed algorithms.
result Achieves superior robustness guarantees and near-optimal sample complexities in both offline and online settings.
New findings show privacy affects generalization error in a non-monotonic way.
problem Privacy and robustness in distributed learning.
method Theoretical analysis and matching lower/upper bounds on algorithmic stability.
result Generalization error is non-monotonically affected by privacy, depending on noise level.
Paper proposes a secure protocol for federated learning.
problem Combining robustness, privacy, and security in federated learning.
method Secure two-server protocol for federated learning.
result Offers both input privacy and Byzantine-robustness.
Paper improves statistical efficiency of median-of-means estimator for Byzantine robust distributed inference.
problem Byzantine robustness in distributed learning systems.
method Variance reduced median-of-means (VRMOM) estimator for Byzantine robust distributed inference.
result Achieves a fast convergence rate with only a constant number of rounds of communications.
A new method balances model quality and Byzantine robustness in Federated Learning.
problem Byzantine clients sending arbitrary or malicious information.
method Practical weight-truncation-based preprocessing method.
result Empirically demonstrates good balance between model quality and Byzantine robustness.
New method improves model accuracy in Byzantine-robust distributed learning by optimizing batch size.
problem Reduces model accuracy drop due to large variance of stochastic gradients in Byzantine-robust distributed learning.
method Proposes ByzSGDnm, a novel BRDL method that uses normalized momentum to mitigate accuracy drop in large batch sizes.
result The optimal batch size increases with the fraction of Byzantine workers, leading to better model accuracy under Byzantine attacks.
Paper develops efficient algorithms for robust distributed learning with statistical guarantees.
problem Limited communication power and adversarial node behaviors in distributed learning.
method Surrogate likelihood framework and median/trimmed mean operations.
result Provable robustness against Byzantine failures and optimal statistical rates.
DynBRO learns robustly from dynamic Byzantine workers.
problem Fault-tolerant distributed learning with dynamic Byzantine workers.
method Multi-level Monte Carlo (MLMC) gradient estimation and adaptive learning rate.
result DynaBRO nearly matches static setting's convergence rate with O ( T ) \mathcal{O}(\sqrt{T}) O ( T ) Byzantine worker changes. Paper develops Byzantine-resilient algorithms for decentralized learning.
problem Vulnerability of distributed learning to Byzantine attacks.
method Dual approach for decentralized optimization.
result Convergence guarantees and experimental validation of the proposed algorithm.
New federated learning protocols resist Byzantine failures and offer privacy guarantees.
problem Resisting Byzantine failures in federated learning.
method Proposes robust federated learning protocols with optimal statistical rates and privacy guarantees.
result Achieves nearly optimal statistical rates and tight rate in terms of all parameters for strongly convex losses.
New algorithm improves decentralized learning in the presence of Byzantine faults.
problem Byzantine faults in decentralized learning on arbitrary graphs.
method Proposes ClippedGossip for Byzantine-robust consensus and optimization.
result First to provably converge to a specified neighborhood of the stationary point for non-convex objectives.
Paper tackles Byzantine attacks in distributed learning with a new ADMM method.
problem Byzantine workers sending arbitrary messages bias distributed learning.
method Byzantine-robust stochastic ADMM exploiting separable problem structure.
result Proposed method converges to optimal solution at O(1/k) rate.
Unified framework for Byzantine robust gossip algorithms with guaranteed performance.
problem Vulnerability of decentralized machine learning to misbehaving devices.
method Introduces F-RG framework and CS+ robust aggregation rule for Byzantine resilience.
result CS+-RG has near-optimal breakdown tolerance and outperforms existing methods.
A method to robustly federate learning with non-i.i.d. data and Byzantine workers.
problem Byzantine workers sending malicious messages in federated learning with non-i.i.d. data.
method Resampling strategy to reduce inner and outer variation, stochastic average gradient, robust geometric median aggregation.
result The method reaches a neighborhood of the optimal solution at a linear convergence rate and learning error depends on the number of Byzantine workers.
Proposes a method to improve Byzantine-robustness in compressed federated learning.
problem Byzantine-robustness in compressed federated learning.
method Gradient difference compression and stochastic average gradient algorithm (SAGA).
result The proposed method reaches a neighborhood of the optimal solution at a linear convergence rate.
Paper tackles Byzantine attacks in Federated Learning by clustering and robustifying.
problem Adversarial attacks from Byzantine machines in Federated Learning.
method Iterative Federated Clustering Algorithm (IFCA) with trimmed mean and median aggregation.
result Improved convergence rate for strongly convex loss functions in Byzantine-Robust IFCA.
A new Federated Learning approach balances personalization and global training.
problem Breaking the curse of data heterogeneity in Federated Learning.
method Splitting variables into global and local parameters, using a simple algorithm.
result The approach allows each client to fit their data perfectly, breaking the curse of data heterogeneity.
In large-scale distributed learning, security issues have become increasingly important. Particularly in a decentralized environment, some computing units may behave abnormally, or even exhibit Byzantine failures -- arbitrary and potentially adversarial behavior. In this paper, we develop distributed learning algorithm…
We study robust distributed learning that involves minimizing a non-convex loss function with saddle points. We consider the Byzantine setting where some worker machines have abnormal or even arbitrary and adversarial behavior. In this setting, the Byzantine machines may create fake local minima near a saddle point tha…
We study a recently proposed large-scale distributed learning paradigm, namely Federated Learning, where the worker machines are end users' own devices. Statistical and computational challenges arise in Federated Learning particularly in the presence of heterogeneous data distribution (i.e., data points on different de…
Paper addresses robust federated linear bandits against Byzantine attacks.
problem Byzantine attacks on a small fraction of agents in federated learning.
method Proposes a geometric median-based robust aggregation oracle.
result Achieves sublinear regret bound of i l d e O ( T 3 / 4 ) ilde{\mathcal{O}}({T^{3/4}}) i l d e O ( T 3/4 ) robust to fewer than half Byzantine agents. We develop a communication-efficient distributed learning algorithm that is robust against Byzantine worker machines. We propose and analyze a distributed gradient-descent algorithm that performs a simple thresholding based on gradient norms to mitigate Byzantine failures. We show the (statistical) error-rate of our al…
Paper addresses Byzantine attacks in decentralized optimization over networks.
problem Byzantine attacks in decentralized stochastic optimization over static and time-varying networks.
method Formulate a TV norm-penalized approximation of the problem, solve using stochastic subgradient method.
result Proposed method reaches a neighborhood of the Byzantine-free optimal solution.
Federated learning enables training collaborative machine learning models at scale with many participants whilst preserving the privacy of their datasets. Standard federated learning techniques are vulnerable to Byzantine failures, biased local datasets, and poisoning attacks. In this paper we introduce Adaptive Federa…
Paper introduces a new distributional successor measure for reinforcement learning.
problem Learning the distributional consequences of behavior in reinforcement learning.
method Formulates distributional successor measure as a distribution over distributions, proposes algorithm to learn it from data.
result Demonstrates zero-shot risk-sensitive policy evaluation.
Study on distributional TD learning with linear approximations for better return estimation.
problem Estimating the return distribution of a policy in reinforcement learning.
method Finite-sample analysis of distributional TD learning with linear function approximation, using the linear-categorical Bellman equation and exponential stability arguments for products of random matrices.
result Sample complexity of linear distributional TD learning matches that of classic linear TD learning, indicating similar difficulty in estimating return distribution versus its expectation.
Paper settles sample complexity for learning from multiple distributions.
problem Learning from multiple data distributions with a hypothesis class of bounded VC dimension.
method Introduced an algorithm with sample complexity of O((d+k)ε^-2)·(k/ε)^o(1).
result Algorithm matches lower bound up to sub-polynomial factor.
New algorithms improve distributional TD learning with linear approximations.
problem Estimating return distributions in reinforcement learning.
method Fine-grained analysis of linear-categorical Bellman equation, variance reduction techniques.
result Tight sample complexity bounds for distributional TD learning with linear approximations.
Paper develops a new algorithm for distribution regression with optimal learning rates.
problem Distribution regression with limited second-stage samples.
method Multi-penalty regularization in a reproducing kernel Hilbert space.
result Derives optimal learning rates for distribution regression.
Introduces NQ network for non-crossing quantile learning.
problem Quantile crossing issue in distributional learning.
method Non-negative activation functions ensure monotonic distributions.
result Effective for distributional reinforcement learning and causal effect estimation.
Unsupervised learning representations generalize better than supervised learning under distribution shifts.
problem Robustness of unsupervised representations to distribution shift.
method Extensive evaluation on synthetic and realistic datasets, including controllable domain generalization datasets.
result Unsupervised representations learned from SSL and AE generalize better than supervised learning under various distribution shifts.
Method learns statistics of return distributions via neural networks and maximum mean discrepancy.
problem Learning probability distributions in reinforcement learning.
method Maximum mean discrepancy (MMD) for learning unrestricted statistics of return distributions.
result Method outperforms standard distributional RL baselines on Atari games.
We provide new results for noise-tolerant and sample-efficient learning algorithms under s s s -concave distributions. The new class of s s s -concave distributions is a broad and natural generalization of log-concavity, and includes many important additional distributions, e.g., the Pareto distribution and t t t -distribution.…
The paper explores solutions to the distributional Bellman equation in reinforcement learning.
problem Distributional reinforcement learning considers complete return distributions, not just expected returns.
method Study existence and uniqueness of solutions to general distributional Bellman equations, linking them to multivariate affine equations.
result Any solution to a distributional Bellman equation can be derived from a multivariate affine distributional equation.
New approach for large-scale distributed learning systems that improve generalization performance.
problem Transitioning from centralized to distributed AI systems for complex learning tasks.
method Self-organizing hierarchical structuring mechanism based on agglomerative clustering, hierarchical generalization, and personalized learning.
result Demonstrates better generalization performance compared to conventional federated learning algorithms.
Paper proposes a new method to learn distribution kernels via entropy maximization.
problem Challenges in applying kernel methods to distribution regression tasks.
method Proposes a novel objective for unsupervised learning of data-dependent distribution kernels based on entropy maximization.
result Demonstrates the effectiveness of the learned kernel across different modalities.
Paper optimizes training data distribution for better model performance across various deployment conditions.
problem Improving model accuracy when deployed with parameters far from training data.
method Developed adaptive algorithms based on bilevel or alternating optimization in the space of probability measures.
result Optimized training distributions lead to models with improved sample complexity and robustness to distribution shift.
Equivariant flows learn symmetrical distributions on manifolds.
problem Learning symmetrical distributions on arbitrary manifolds.
method Equivariant manifold flows.
result Learned gauge invariant densities over SU(n) in quantum field theory.
Distributed machine learning algorithms enable learning of models from datasets that are distributed over a network without gathering the data at a centralized location. While efficient distributed algorithms have been developed under the assumption of faultless networks, failures that can render these algorithms nonfu…
Method enhances anomaly detection using contrastive learning and out-of-distribution data.
problem Improving anomaly detection in datasets with limited out-of-distribution data.
method Proposes a contrastive learning method that incorporates out-of-distribution data to enhance anomaly detection performance.
result The method significantly improves anomaly detection performance, even with limited out-of-distribution data.
New bounds on learning from multiple distributions for VC classes.
problem Understanding the sample complexity of learning from multiple data distributions.
method Analyzing the gap between known upper and lower bounds for PAC-learnable classes.
result Recent progress on sample complexity for VC dimension d classes on k distributions.
Proposes MRO to achieve uniformly low regret in distributionally robust learning.
problem Learning under unknown test distributions (distribution shift).
method Minimax Regret Optimization (MRO) for robust machine learning.
result MRO achieves uniformly low regret across all test distributions.
New perspective on distribution shift helps make learning easier.
problem Learning with different training and target distributions.
method Formalizing and exploring Positive Distribution Shift (PDS).
result Distribution shift can be positive, making learning easier.
Theory of learning with weight-distribution constraints.
problem Understanding how structure influences function in neural networks.
method Statistical mechanical theory and optimal transport.
result Reduction in capacity due to constrained weight-distribution is related to Wasserstein distance.
New RL algorithm minimizes distributional learning error.
problem Improving distributional reinforcement learning for better error minimization.
method Proposes a new model-based algorithm with theoretical minimax optimality.
result Proves minimax optimality for approximating return distributions.
In this paper we argue for the fundamental importance of the value distribution: the distribution of the random return received by a reinforcement learning agent. This is in contrast to the common approach to reinforcement learning which models the expectation of this return, or value. Although there is an established …