Robuta

With the advent of large datasets, offline reinforcement learning (ORL) is a promising framework for learning good decision-making policies without interacting...

A theoretically grounded new algorithm for imitation learning.

Imitation Learning (IL) enables agents to mimic expert behavior by learning from demonstrations. However, traditional IL methods require large amounts of...

05/16/19 - We consider the problem of imitation learning from a finite set of expert trajectories, without access to reinforcement signals. T...

Imitation Learning (IL) and Reinforcement Learning (RL) are often introduced assimilar, but separate problems. Imitation learning involves a supervisor thatp...

In unmanned systems, remote sensing is an approach that collects and analyzes data such as visual images, infrared thermal images, and LiDAR sensor data from a...

We propose Constraint-Aware Diffusion Guidance (CoDiG), a constraint-aware imitation learning framework based on conditional diffusion models. Unlike...

This paper provides the first expert sample complexity characterization for learning a Nash equilibrium from expert data in Markov Games. We show that a new...

Most deep learning methods attempt to learn artificial neural networks from scratch, using architectures or neurons or approaches often only very...