Reproducibility, Reusability, & Robustness in Deep Reinforcement Learning - Prof. Pineau

Показать описание

Recorded May 3rd, 2018
"In recent years, significant progress has been made in solving challenging problems across various domains using deep reinforcement learning. However reproducing results for state-of-the-art deep RL methods is seldom straightforward. High variance of some methods can make learning particularly difficult when environments or rewards are strongly stochastic. Furthermore, results can be brittle to even minor perturbations in the domain or experimental procedure. In this talk, I will discuss challenges that arise in experimental techniques and reporting procedures in deep RL, and will suggest methods and guidelines to make future results more reproducible, reusable and robust. I will also report on findings from the ICLR 2018 reproducibility challenge." - Prof. Joelle Pineau

Рекомендации по теме

Комментарии

"Solving a problem" - eg, getting a goal in football, should be as abstract, general and generic as possible ("getting a goal"). So long as the system can produce many different solutions to achieve the same objective, then the goal "solving a problem" has been successfully reproduced, repeatedly, even though the path to it are different.

tthtlc

Creativity == Something different, and thus not necessarily identical to past pattern. Non-reproducibility is a sign of creativity.

tthtlc

Not to oversimplify Learning Like shocking a dog or providing a treat?

brianvandenberg

i dont know why, but i cant listen to female presenters. its like they dont say anything of substance, my brain is simply incapable of understanding what is it they say. i listent to 13 minutes of this and im done

lucid

Reproducibility, Reusability, & Robustness in Deep Reinforcement Learning - Prof. Pineau

Reproducibility, Reusability, & Robustness in Deep Reinforcement Learning - Prof. Pineau

Joelle Pineau: Reproducibility, Reusability, and Robustness in Deep Reinforcement Learning ICLR 2018

15. Reproducibility, Reusability, And Robustness In Deep Reinforcement Learning

Dr. Joelle Pineau - Reproducible, Reusable, and Robust Reinforcement Learning - NeurIPS 2018

Reproducible, Reusable, and Robust Reinforcement Learning -Joelle Pineau

Reproducibility in Data Science | NYU

Building Reproducible, Reusable & Robust Deep Reinforcement Learning Systems | Jöelle Pineau

MLOps: From notebooks to reproducibility and reusability (Part 1: ML Development and experiments)

ACM DEBS 2020: Building Reproducible Reusable and Robust Machine Learning Software (Joelle Pineau)

Reproducibility, Replicability, and Robustness: J. Flanagan LADAL Opening Webinar Series 2021

Webinar: Best Practices for Reproducible, Collaborative Data Science

TechAide AI4Good 2020 - Joelle Pineau: Building Reproducible, Reusable and Robust Deep RL Systems

Reusedata: An Open-Source, Open-Development Tool For Reusable & Reproducible Genomic Data Manage...

Enable Reproducible, Scalable, and Orchestrated Workflows in HPC and Cloud-Converged Infrastructure

Week 4 - Reproducibility in Machine Learning: From Theory to Practice - Koustuv Sinha

Reproducibility in High Performance Computing

Open Science, Reuse, and Reproducibility with Flywheel Webinar

Implementing Increased Transparency and Reproducibility in Economics — Lars Vilhuber

Aidan Heerdegen: RRR: Reliability, Replicability, Reproducibility for climate models

Checking a reproducible data analysis using make, git, and R: Testing with new data! (CC075)

Reproducible and Open Data Science

METASCIENCE 2021 9 23 Building a Research Culture for Replicability and Reproducibility in the...

Implementing Increased Transparency and Reproducibility in Economics

Is AI Facing A Reproducibility Crisis?