Simon Schug's starred repositories
recurrentgemma
Open weights language model from Google DeepMind, based on Griffin.
minimal-hypernetwork
Minimal hypernetwork implementation.
modular-hyperteacher
Code accompanying the paper Discovering modular solutions that generalize compositionally
machine-tasks
Datasets for compositional learning
groundedSCAN
Grounded SCAN data set.
shrink-perturb
Optax implementation of shrink and perturb (Ash & Adams, 2020).
permuted_initialisations
Code accompanying "Random initialisations performing above chance and how to find them", for paper see https://arxiv.org/abs/2209.07509
amortized-optimization-tutorial
Tutorial on amortized optimization for learning to optimize over continuous domains
jax-hypernetwork
A simple hypernetwork implementation in jax using haiku.
gymnax-blines
Baselines for gymnax 🤖
eprop-PyTorch
PyTorch implementation of the eligibility propagation (e-prop) learning algorithm.
BinomialSynapses.jl
High-performance Julia package for online learning of binomial synaptic release models
minimal-meta-rl
Minimal A2C/A3C example of an LSTM-based meta-learner.
presynaptic-stochasticity
Presynaptic Stochasticity Improves Energy Efficiency and Alleviates the Stability-Plasticity Dilemma
hypnettorch
Package for working with hypernetworks in PyTorch.
deep_feedback_control
Official code repository for the publication "Credit Assignment in Neural Networks through Deep Feedback Control".