×
Dec 20, 2018 · The main approach, Universal Successor Features Approximators (USFAs) is a combination of two recent approaches: Universal Value Function ...
Our proposed universal successor features approximators (USFAs) ... We call the resulting model universal successor features approximators, or USFAs for short.
We present an improved version of Universal Successor Features based DRL method which can improve the transfer learning of agents.
Another generalization of value functions that is related to SFs is Schaul et al.'s [20] universal value function approximators (UVFAs). UVFAs extend the ...
(4 seeds). Baselines. (1) Universal Successor. Feature Approximators (USFA) [8] is the only method that shares an SF estimator across tasks ...
In this paper, we propose (1) Universal Successor Features (USFs) to capture the underlying dynamics of the environment while allowing generalization to unseen ...
The main idea is to to construct a single function approximator V (s; θ) that estimates the long-term reward from any state s, using parameters θ.
In this work, we propose the "Successor Features Keyboard" (SFK), which enables transfer with discovered state-features and task encodings.
It is clear that C = {w0} recovers a universal value function approximator, minimising the value function approximation error. Conversely, if. C = M, it means ...
Vusfa: Variational universal suc- cessor features approximator to improve transfer drl for target driven visual navigation. arXiv preprint. arXiv:1908.06376, ...