🛬 Vancouver for #icml, I’ll be presenting our work on Subgoal Guided Heuristic Search with Learned Subgoals on Tuesday from 4:30-7:00pm. Come stop by and say hello 👋

🧵1/ New paper! 📄 Subgoal-Guided Policy Heuristic Search with Learned Subgoals, led by my PhD student @tuero.ca.

arxiv.org/pdf/2506.07255

This paper follows the Levin tree search (LTS) research line and focuses on learning subgoal-based policies.

This enables our method to learn from failed attempts where we do not solve the problem outright, but solve some of the subgoals. The results show that this approach helps the system learn faster and more efficiently, without sacrificing the quality of the policy.

This paper looks at a class of algorithms called policy tree search, which combines policies from reinforcement learning with traditional tree search. We show how one can decompose a problem into learnable subgoals, without any prior knowledge of the environment.

Excited to share our paper "Subgoal-Guided Policy Heuristic Search with Learned Subgoals" has been accepted to #icml2025!

Paper Preview: arxiv.org/pdf/2506.07255

I'll be attending ICML-25 in Vancouver, and I'm looking forward to chatting with anyone who is interested in our work!

GitHub - tuero/tinytensor: A C++ multi-dimensional tensor library with automatic gradient tracking and CUDA acceleration. A C++ multi-dimensional tensor library with automatic gradient tracking and CUDA acceleration. - tuero/tinytensor

I implemented a pytorch-like cuda accelerated autograd library in C++ to learn cuda and the performance pain points when using these frameworks. Check it out! github.com/tuero/tinyte... #cpp #cplusplus #cuda #MLSky #DeepLearning

Sometimes it’s beneficial to try and reformulate a problem into another so that you can lean on the solutions the other provides. A Conv2D can be rewritten as a series of matrix multiplications, and I’ve gained a 2x speed up by reusing my fast matmul kernels for my Conv2D implementation!

Thanks for these lists! Is the grumpy list an inside joke or am I missing something 🤣

So much for sleep, I’m glued here finding everyone to follow!