Releases: patham9/NACE
Releases · patham9/NACE
v0.0.9
With this release, the most important objectives of this POC have been accomplished.
Features:
- This release makes the agent's behavior even more rational.
- While some poorly-understood issues remain (leaving room for future improvements), this agent is a highly capable causal learner.
- A new visualization is included from which the demonstration gifs were automatically exported.
v0.0.8
v0.0.8
- First time that the soccer and egg level work in most of the cases too.
- Rejection sampling (a Montle Carlo method) for non-deterministic filtering of hypotheses, the remaining considered set tends to include high truth-expectation candidates, which makes the overall system more robust to failed associations and in case the prediction-observation mismatches do not resolve relevant prediction errors.
TODO next: higher robustness to externally induced changes.