Blog über „Inside the Matrix: Visualizing Matrix Multiplication, Attention and Beyond“:

https://pytorch.org/blog/inside-the-matrix/

Hier eine Veröffentlichung über: „MiniZero: Comparative Analysis of AlphaZero and MuZero on Go, Othello, and Atari Games“:

https://arxiv.org/pdf/2310.11305.pdf

Veröffentlichung über „MiniZero: Comparative Analysis of AlphaZero and MuZero on Go, Othello, and Atari Games“:

https://arxiv.org/pdf/2310.11305.pdf

Veröffentlichung über „Student of Games: A unified learning algorithm for both perfect and imperfect information games“:

https://www.science.org/doi/10.1126/sciadv.adg3256

Hier eine grundsätzliche Veröffentlichung über Transformer:

„Decision Transformer: Reinforcement Learning via Sequence Modeling“.

https://arxiv.org/pdf/2106.01345.pdf

Evidence of Learned Look-Ahead in a Chess-Playing Neural Network

https://arxiv.org/pdf/2406.00877

Sehr interessante Veröffentlichung.

## Representation Matters for Mastering Chess: Improved Feature Representation in AlphaZero Outperforms Switching to Transformers

<https://www.aiml.informatik.tu-darmstadt.de/papers/czech24representation.pdf>

From 20 seconds skimming through, they claim that adding those planes to standard resnet planes outperform (some definition of) switching to transformers:

* grouped mask of all P1 pieces
* grouped mask of all P2 pieces
* Checkerboard
* 5 planes filled with the same value: number of {PAWN, KNIGHT, BISHOP, ROOK, QUEEN} we have
* 5 planes filled with the same value: diff of {PAWN, KNIGHT, BISHOP, ROOK, QUEEN} "we" and "opponent" has
* All ones if there are opposite color bishops
* all pieces giving check
+150 Elo

Interessante Veröffentlichung über KI-Self-Learning im Schach:

https://arxiv.org/pdf/2402.04494