hardmaru (@hardmaru) on X

X (formerly Twitter) ·

1 min read Original article ↗

@hardmaru

Minimum Description Length Recurrent Neural Networks Neural networks trained to optimize their Minimum Description Length, e.g. balance between the network’s complexity and its accuracy at a task, results in small networks with transparent inner workings. arxiv.org/abs/2111.00600

3:41 AM · Apr 5, 2022