New algorithms for learning under s-concave distributions, including Pareto and t-distributions.
problem Learning under broad and natural generalizations of log-concave distributions, including fat-tailed ones.
method Introduce new convex geometry tools to study s-concave distributions and use these properties to provide bounds on learning quantities. result Significantly generalize prior results for margin-based, disagreement-based, and passive learning of intersections of halfspaces.
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.
The paper emphasizes the importance of value distribution in reinforcement learning.
problem The focus on value expectation in reinforcement learning is insufficient.
method Developed a new algorithm based on the distributional perspective of reinforcement learning.
result Demonstrated significant distributional instability in reinforcement learning control.
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.
JointGAN learns joint distributions across multiple domains.
problem Learning joint distributions across multiple domains.
method Generative adversarial nets (GANs) with multiple generators and a critic, jointly trained.
result Synthesis of draws from full joint distribution.
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.
Paper analyzes CDRL algorithms for reinforcement learning.
problem Understanding theoretical properties of CDRL algorithms.
method Introduces a framework to analyze CDRL algorithms, establishes the importance of the projected distributional Bellman operator, draws connections to Cramér distance, and proves convergence.
result Proof of convergence for sample-based categorical distributional reinforcement learning algorithms.
ByRDiE algorithm handles Byzantine failures in decentralized learning.
problem Byzantine failures in distributed learning.
method Byzantine-resilient distributed coordinate descent (ByRDiE) algorithm.
result ByRDiE enables high-dimensional distributed learning in the presence of Byzantine failures.
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.
D4PG combines distributional reinforcement learning with distributed learning for control tasks.
problem Continuous control tasks in reinforcement learning.
method Adapting distributional reinforcement learning to continuous control, using a distributed framework, N-step returns, and prioritized experience replay.
result D4PG achieves state-of-the-art performance across various control tasks.
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.
Unified approach to stabilize adversarial learning for joint distribution matching.
problem Non-identifiability issues in bidirectional adversarial training.
method Unified framework of adversarial and non-adversarial approaches, stabilizing learning.
result Stabilized learning of unsupervised and semi-supervised bidirectional adversarial methods.
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.
Paper analyzes distributed learning with non-i.i.d. samples.
problem Learning rate analysis for distributed kernel ridge regression with dependent samples.
method Integral operator approach and covariance inequality for strong mixing sequences.
result Derives optimal learning rates for distributed kernel ridge regression.
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.
A new distributed learning method for high-dimensional linear classification.
problem Efficiently performing linear classification on large-scale, high-dimensional data.
method Feature-distributed stochastic variance reduced gradient (FD-SVRG) for high-dimensional linear classification.
result FD-SVRG outperforms other distributed methods in terms of communication cost and wall-clock time.
Paper proposes DDA for better transfer learning performance.
problem Domain discrepancy between source and target distributions.
method Dynamic Distribution Adaptation (DDA) to evaluate and adapt distribution importance.
result DDA improves transfer learning performance on various tasks.
GAN Q-learning uses GANs for distributional RL in tabular and gym environments.
problem Complex MDPs in nonlinear function approximation.
method Generative adversarial networks (GANs) for distributional reinforcement learning.
result Empirically shows GAN Q-learning is a viable alternative to traditional methods.
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.
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 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.
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.
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.
A new method for reinforcement learning using quantile regression.
problem Improving reinforcement learning algorithms for better performance.
method Using quantile regression to approximate the full quantile function for state-action return distributions.
result Improved performance on Atari games, demonstrating the effectiveness of the method.
A new algorithm for distributed machine learning with observations and features.
problem Optimization in large-scale distributed machine learning.
method A stochastic algorithm for distributed observations and features with convergence analysis.
result The algorithm outperforms a benchmark in early iterations.
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.
A new autoencoder learns expressive posterior and conditional likelihood distributions.
problem Learning more expressive posterior and conditional likelihood distributions.
method Implicit autoencoder using two generative adversarial networks for reconstruction and regularization.
result Implicit autoencoder can disentangle content and style information.
Secure and efficient distributed learning on devices with limited communication.
problem Limited communication and security in distributed on-device learning.
method Proposes SLSGD, a robust distributed optimization algorithm with efficient communication and attack tolerance.
result Stabilizes convergence and tolerates data poisoning on a small number of workers.
Proposes BDA for better transfer learning performance.
problem Distribution divergence between source and target domains, especially marginal and conditional.
method Balanced Distribution Adaptation (BDA) and Weighted BDA (W-BDA) algorithms.
result Improves transfer learning performance on both balanced and imbalanced datasets.
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.
TonY simplifies distributed ML job management.
problem Managing distributed ML jobs is complex and resource-intensive.
method TonY is an open-source orchestrator for distributed ML jobs.
result TonY simplifies distributed ML job management.
New findings on PAC learning and marginal distribution estimation.
problem Understanding how PAC learning relates to marginal distribution estimation under distributional constraints.
method Revisited the connection between PAC learning, uniform convergence, and density estimation, considering a known family of marginal distributions.
result PAC learning is sandwiched between two refined models of density estimation, differing only in whether the learner knows the set of well-estimated events in H.
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.
New algorithms improve label complexity for active multi-distribution learning.
problem Active multi-distribution learning with improved label complexity.
method Developed new algorithms for active multi-distribution learning and established improved label complexity upper and lower bounds.
result Improved label complexity upper and lower bounds for active multi-distribution learning.
NC learns all conditional distributions of a random vector.
problem Learning all conditional distributions of a random vector.
method Adversarial training to match each conditional distribution.
result NC generalizes to sample from conditional distributions never seen.
A new method for learning conditional distributions using ODEs and neural networks.
problem Learning conditional distributions efficiently and accurately.
method Conditional Föllmer Flow, discretized with Euler's method, using nonparametric velocity estimation.
result Effective approximation of target conditional distributions, with convergence results for Wasserstein-2 distance.
The paper proposes noise-invariant distances and features for robust testing and learning.
problem Testing and learning on distributions with irrelevant noise.
method Kernel embeddings, Maximum Mean Discrepancy, distances invariant to additive symmetric noise.
result Noise-invariant distances and features for robust testing and learning.
Overview of distributed learning algorithms for finitely many hypotheses.
problem Distributed learning with finitely many hypotheses.
method Various approaches to distributed learning, including social learning algorithms.
result Convergence and convergence rate results for both asymptotic and finite time regimes.
Paper introduces a new method for learning with distributions using dissimilarity measures.
problem Learning with probability distributions using dissimilarity measures.
method Introduces embeddings based on dissimilarity of distributions to templates, extending similarity theory to population distributions.
result Proves that dissimilarity theory holds for empirical distributions and shows better performance of Wasserstein distance embedding.
This paper improves reinforcement learning by modeling return distributions explicitly.
problem Traditional reinforcement learning averages over randomness, missing distributional information.
method Develops a distributional reinforcement learning algorithm using quantile regression.
result The new algorithm significantly outperforms existing methods on Atari games.
Distributed learning for related tasks using graph weights.
problem Learning multiple tasks on different machines.
method Weighted averaging of messages with skewing or stepsize control.
result Different tasks can be learned on different machines.
Distribution networks model novel classes in open set learning.
problem Modeling novel classes in open set learning.
method Distribution networks map samples to a latent space where known and novel classes' distributions are jointly learned.
result Distribution networks accurately detect and model novel classes for subsequent classification.
The theory of learning under the uniform distribution is rich and deep, with connections to cryptography, computational complexity, and the analysis of boolean functions to name a few areas. This theory however is very limited due to the fact that the uniform distribution and the corresponding Fourier basis are rarely …
SHIFT method optimally estimates heterogeneous discrete distributions with limited communication.
problem Collaborative learning of discrete distributions under heterogeneity and communication constraints.
method Two-stage method: First, users learn a central distribution; then, fine-tune this to estimate individual distributions.
result SHIFT is minimax optimal in the model of heterogeneity and under communication constraints.