Understanding the Effects of Dataset Characteristics on Offline Reinforcement Learning

2 code implementations • 8 Nov 2021 • Kajetan Schweighofer, Markus Hofmarcher, Marius-Constantin Dinu, Philipp Renz, Angela Bitto-Nemling, Vihang Patil, Sepp Hochreiter

Algorithms that constrain the learned policy towards the given dataset perform well for datasets with high TQ or SACo.

Offline RL reinforcement-learning

CLOOB: Modern Hopfield Networks with InfoLOOB Outperform CLIP

1 code implementation • 21 Oct 2021 • Andreas Fürst, Elisabeth Rumetshofer, Johannes Lehner, Viet Tran, Fei Tang, Hubert Ramsauer, David Kreil, Michael Kopp, Günter Klambauer, Angela Bitto-Nemling, Sepp Hochreiter

CLOOB consistently outperforms CLIP at zero-shot transfer learning across all considered architectures and datasets.

Contrastive Learning Self-Supervised Learning +1

Hopular: Modern Hopfield Networks for Tabular Data

no code implementations • 29 Sep 2021 • Bernhard Schäfl, Lukas Gruber, Angela Bitto-Nemling, Sepp Hochreiter

In experiments on small-sized tabular datasets with less than 1, 000 samples, Hopular surpasses Gradient Boosting, Random Forests, SVMs, and in particular several Deep Learning methods.