The Feynman Machine: AI architecture in Clojure/Neanderthal
summary
Implements the Feynman Machine in Clojure with two runtimes: an OpenCL/GPU path (AugmentEO) and a CPU C++ library with JNI wrappers, plus idiomatic Clojure wrappers (~100–200 lines) generated with SWIG. Uses Neanderthal for near‑native linear algebra, sparse binary vectors and a winner‑takes‑all encoding, trains from local prediction errors (no backprop). Demonstrates a self‑driving on Raspberry Pi prototype that learns the driving task in about two to three minutes; several live demos and working wrappers are available.
AI-written summary · about
chapters (4)
related talks
25mMore Clojure Less Complication
41mInteractive GPU programming with ClojureCUDA & ClojureCL
31mScientific Computing with Clojure
42mClojure is Not Afraid of the GPU
29mProgramming machine learning algorithms in hardware, sanely, using Haskell and Rust!
29mDeep Learning Needs Clojure
40mBuilding Machine Learning Models with Clojure and Cortex
36mLearning Clojure and CLJS by playing a card game