Work done as a visiting student at MIT. 38th Conference on Neural Information Processing Systems (NeurIPS 2024).

Invariant Risk Minimization (IRM) is a recently proposed framework for outof-distribution (o.o.d) generalization.

We demonstrate the efficiency of our protocol for distributed optimization with strongtheoreticalguarantees.

Nevertheless, theperformance guarantee of adaptive TD with neural network approximation remains widely unknown.

The impressive capability of these models, however, often entails significant computational overhead during both training and inference.