Schmidhuber
A computer scientist known for work on recurrent neural networks and linear Transformers.
Timeline (last 90 days)
The concept of fast weights can be traced back to the fast weight associative memory proposed by Hinton and von der Malsburg in the 1990s, as well as Schmidhuber's linear Transformer work.
Mattick proposes a distinction: if data collection is inside the loop, it is model-based RL; if outside the loop, it is a world model; Schmidhuber's original world model paper is essentially model-based reinforcement learning.
The world model concept was systematically proposed by Ha and Schmidhuber in their 2018 paper "World Models".
All Facts (3)
The concept of fast weights can be traced back to the fast weight associative memory proposed by Hinton and von der Malsburg in the 1990s, as well as Schmidhuber's linear Transformer work.
50%UnverifiedMattick proposes a distinction: if data collection is inside the loop, it is model-based RL; if outside the loop, it is a world model; Schmidhuber's original world model paper is essentially model-based reinforcement learning.
50%UnverifiedThe world model concept was systematically proposed by Ha and Schmidhuber in their 2018 paper "World Models".
50%