Fabio Viola

Title: Fabio Viola: Inventor in Reinforcement Learning

Introduction

Fabio Viola is an inventor based in Montreal, Canada. His patent work concerns reinforcement learning, including methods for helping an agent choose actions while interacting with an environment.

Latest Patents

Viola is associated with 1 patent: “Training Action Selection Neural Networks Using Hindsight Modelling.” The invention describes a reinforcement learning method and system that selects actions for an agent interacting with an environment. A hindsight-model neural network learns a causal model from future environment-state trajectories. Because future information is unavailable when the agent selects an action, this hindsight model trains a separate model neural network using current observations. The latter network learns to predict the hindsight model’s output, supporting action selection based on information available at the time.

Career Highlights

The provided information associates Viola with GDM Holding LLC. His patent reflects work on neural-network training and action selection in reinforcement learning.

Collaborations

The listed coworkers are Arthur Clement Guez and Theophane Guillaume Weber. Their names are provided as collaborators associated with Viola’s professional context.

Conclusion

Fabio Viola’s identified patent focuses on using hindsight-based training to improve action selection in reinforcement learning, connecting predictive neural-network methods with an agent’s present observations.