↓Skip to main content

🏢 McGill University

Trajectory Flow Matching with Applications to Clinical Time Series Modelling
·1814 words·9 mins· loading · loading
AI Applications Healthcare 🏢 McGill University
Simulation-free Neural SDE training via Trajectory Flow Matching unlocks scalability and stability for modeling complex real-world time series, particularly in clinical settings.
The High Line: Exact Risk and Learning Rate Curves of Stochastic Adaptive Learning Rate Algorithms
·1819 words·9 mins· loading · loading
Machine Learning Optimization 🏢 McGill University
Researchers developed a framework for analyzing stochastic adaptive learning rate algorithms, providing exact risk and learning rate curves, revealing the importance of data covariance and uncovering …
Robot Policy Learning with Temporal Optimal Transport Reward
·2204 words·11 mins· loading · loading
Machine Learning Reinforcement Learning 🏢 McGill University
Temporal Optimal Transport (TemporalOT) reward enhances robot policy learning by incorporating temporal order information into Optimal Transport (OT)-based proxy rewards, leading to improved accuracy …
Periodic agent-state based Q-learning for POMDPs
·2014 words·10 mins· loading · loading
Machine Learning Reinforcement Learning 🏢 McGill University
PASQL, a novel periodic agent-state Q-learning algorithm, significantly improves reinforcement learning in partially observable environments by leveraging non-stationary periodic policies to overcome …
Parseval Regularization for Continual Reinforcement Learning
·2345 words·12 mins· loading · loading
Machine Learning Reinforcement Learning 🏢 McGill University
Boost continual reinforcement learning with Parseval regularization: maintaining orthogonal weight matrices preserves optimization, significantly improving RL agent training across diverse tasks.
Multivariate Probabilistic Time Series Forecasting with Correlated Errors
·6695 words·32 mins· loading · loading
AI Generated Machine Learning Deep Learning 🏢 McGill University
Boost multivariate time series forecasting accuracy by efficiently learning the complex correlation structure of prediction errors, enhancing reliability without expanding model size.
HardCore Generation: Generating Hard UNSAT Problems for Data Augmentation
·2159 words·11 mins· loading · loading
AI Theory Optimization 🏢 McGill University
HardCore: Fast generation of hard, realistic UNSAT problems for improved SAT solver runtime prediction.
From Linear to Linearizable Optimization: A Novel Framework with Applications to Stationary and Non-stationary DR-submodular Optimization
·1591 words·8 mins· loading · loading
AI Theory Optimization 🏢 McGill University
A novel framework extends optimization algorithms from linear/quadratic functions to a broader class of ‘upper-linearizable’ functions, providing a unified approach for concave and DR-submodular optim…
Efficient Reinforcement Learning by Discovering Neural Pathways
·3467 words·17 mins· loading · loading
Machine Learning Reinforcement Learning 🏢 McGill University
Discover efficient neural pathways for reinforcement learning; drastically reducing model size and energy consumption without sacrificing performance.
Density-based User Representation using Gaussian Process Regression for Multi-interest Personalized Retrieval
·2288 words·11 mins· loading · loading
Machine Learning Recommendation Systems 🏢 McGill University
GPR4DUR leverages Gaussian Process Regression to create density-based user representations for accurate multi-interest personalized retrieval, overcoming limitations of existing methods.
Adaptive Exploration for Data-Efficient General Value Function Evaluations
·2591 words·13 mins· loading · loading
Machine Learning Reinforcement Learning 🏢 McGill University
GVFExplorer: An adaptive behavior policy efficiently learns multiple GVFs by minimizing return variance, optimizing data usage and reducing prediction errors.
Action Gaps and Advantages in Continuous-Time Distributional Reinforcement Learning
·1595 words·8 mins· loading · loading
Machine Learning Reinforcement Learning 🏢 McGill University
Distributional RL’s sensitivity to high-frequency decisions is unveiled, with new algorithms solving existing performance issues in continuous-time RL.
4+3 Phases of Compute-Optimal Neural Scaling Laws
·3282 words·16 mins· loading · loading
AI Theory Optimization 🏢 McGill University
Researchers discovered four distinct compute-optimal phases for training neural networks, offering new predictions for resource-efficient large model training.