World Modelling
Models that learn environment dynamics, latent state, and predictive representations for agents. 4 topics in this category.
intermediate
Joint-Embedding Predictive Architecture
An introduction to how JEPA learns by predicting latent representations, with examples from I-JEPA and action-conditioned world models.
advancedRecurrent State Space Models (RSSM)
An introduction to recurrent state-space models, their training objective, and their evolution through PlaNet, Dreamer, MuDreamer, and R2-Dreamer.
intermediateState Space Models for Sequence Modelling
An introduction to state space models, their continuous and discrete forms, recurrent and convolutional representations, HiPPO, S4, S5, and Mamba.
intermediateWorld Models
Learned environment models that let agents predict, plan, and learn from imagined futures.