Difference between revisions of "Fat Fritz"

From Chessprogramming wiki
Jump to: navigation, search
Line 79: Line 79:
 
[[Category:PC]]
 
[[Category:PC]]
 
[[Category:GPU]]
 
[[Category:GPU]]
[[Category:NN]]
+
[[Category:DCNN]]
 
[[Category:Windows]]
 
[[Category:Windows]]
 
[[Category:Industrial Heritage Trail]]
 
[[Category:Industrial Heritage Trail]]

Revision as of 21:13, 24 August 2020

Home * Engines * Fat Fritz

Unser Fritz [1]

Fat Fritz,
a commercial chess entity by ChessBase featuring a set of custom made neural network weights that work in the open source project Leela Chess Zero within its Lc0 executable. Initially introduced as a cloud project in August 2019, inspired by DeepMind's ground breaking AlphaZero approach combining Deep learning with Monte-Carlo Tree Search (MCTS) [2], and Albert Silver's Deus X experience, relaxing Leela Chess' "Zero" paradigm of pure reinforcement learning due to supervised learning, Fat Fritz was released in November 2019 as part the Fritz 17 package, best supported by a GPU card such as Nvidia GeForce 20 series. The project and its underlying techniques were emphasized by a talk of AlphaZero co-author Thore Graepel at ChessBase, attended by German computer chess icon Thomas Nitsche along with his sons and Math42 authors Raphael Nitsche and Maxim Nitsche [3]. Fat Fritz was further supported by Daniel Uranga [4] [5].

Lc0 and Fat Fritz Diagrams

Lc0diagram.png
Deusx.png

Lc0 and DeusX aka Fat Fritz components [6]

See also

Forum Posts

2019

2020

External Links

ChessBase

2019

2020

Fritz 17

Source Code

Rating Lists

References

Up one Level