🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Last year we showed that deep RL can successfully control stratospheric balloons in the real world go.nature.com/3dtvrPr Today we're announcing a β-release of the BLE, a high-fidelity simulator of this complex real-world decision making problem. bit.ly/3y1i2aK 1/🎈

Life update: I'll be joining the wonderful people in the Intelligent Robot Lab at Brown University this Fall for my PhD!!

A lot of people complain that RL doesn't work and RL researchers are still playing games. While this criticism is true to some extent, there's been a new trend of applying RL for real-life problems. This is a thread of notable papers split by the topic. 1/n

I'm only 50 pages in, but I totally recommend this book. One of the best textbooks I've ever read.

I thought it might be helpful to put together a (not that short) blog post about lessons I've learnt from applying for a (CS) PhD program: link.medium.com/B1k51lBJMpb

Your NeurIPS submission Number: 12*** 😐

Check out our new work! It's an empirical study of the different input augmentations popular in real-world RL. This was work done during my time at the Reinforcement Learning and Artificial Intelligence with two of the best advisors a grad student could ask for - Marlos C. Machado and Adam White.

Our paper has been accepted to Accepted papers at TMLR! This work was led by David Tao while he as a student at University of Alberta with Adam White and I. openreview.net/forum?id=RLYky…

Agent-State Construction with Auxiliary Inputs Ruo Yu Tao, Adam White, Marlos C. Machado. Action editor: Dinesh Jayaraman. openreview.net/forum?id=RLYky… #reinforcement #recurrent #summarizes

Our paper, which was led by David Tao when he was at University of Alberta, is now officially published by TMLR 😁 Shortly: what do most success in deep RL look like in terms of inputs to the network? Why is that so effective? I tweeted about it a while ago: x.com/marloscmachado…

We finally have a version of our paper on loss of plasticity and continual backprop that is polished and submitted to a journal. Good work led by my PhD student Shibhansh Dohare. arxiv.org/abs/2306.13812…

Thrilled to announce the first annual Reinforcement Learning Conference RL_Conference, which will be held at UMass Amherst August 9-12! RLC is the first strongly peer-reviewed RL venue with proceedings, and our call for papers is now available: rl-conference.cc.

For every N likes this post gets, I'll make this Sloth Tensorflow developer more frustrated by the documentation.

Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish colors to hyperparameters for which training diverges.

‼️New preprint (with Tuminauskaite, @Dr_Park_Phd, and Philip E. Tetlock) on the ‘Wisdom of the Silicon Crowd’ in forecasting! We find that: 🔴 LLM crowd forecasting accuracy matches that of a human crowd 🔴 LLM predictions are improved by human input More details and paper below!

We also trained a probe to reconstruct the PacMan dots from the agent’s memory. Guess which agent had an easier time with this… Yep! The λ-discrepancy agent knows where it has been, while the normal RNN agent basically has no idea. 9/n

got my sticker

About time!!! Well deserved to the founders of our field

Imagine being held at gunpoint (bear with me) by a literate animal, and the only hope of rescue is (BEAR WITH ME) tweeting a coded message