Rob van Gastel

dinov3-finetune. Testing adaptation of the DINOv2/3 encoders for vision tasks with Low-Rank Adaptation (LoRA)

meta-rl-algorithms. A collection of Meta-Reinforcement Learning algorithms in PyTorch

rnn-sac. Implementation of the Discrete Soft Actor-Critic algorithm with RNN policy in PyTorch

removing-pos-vit-bias. Using RASA post-training to remove positional bias from pretrained encoders like DINOv3

meta-in-context-learning. Testing the meta in-context learning capabilities of transformers in JAX

svg-priors. Implemenation of the SVG(0) agent with extension of information asymmetric prior in PyTorch

quicksilver. Implementation of an in-memory graph database supporting RPQ queries

Assign. Assign is an app for students to collaborate on their assignments build by Max Wammes and Rob van Gastel

adapt-vjepa-world-model. Can the V-JEPA2 model be used as a world model?

optimization-ml. Testing different properties in optimization of ML algorithms

machine-learning-notes. Tracking read research papers on varying topics in Machine Learning