Skip to content

README.md

Latest commit

 

History

History
36 lines (21 loc) · 788 Bytes

README.md

File metadata and controls

36 lines (21 loc) · 788 Bytes

Diffy

See also: gym

Neural Network for Clojure

  • Dense layer with a single or multiple bias values
  • Sigmoid activation function
  • 3 Matrix implementations:
    • native Clojure vectors
    • clojure.core.matrix
    • uncomplicate.neanderthal.native

The Neanderthal implementation gets enabled by uncommenting the corresponding dependencies in deps.edn and the :require and fixture :each entries in diffy_test.clj.

Tests

$ clj -Atest

Diffy Examples

Mnist

Get mnist_784_csv.csv from https://datahub.io/machine-learning/mnist_784

$ clj -M -m examples.mnist

Uses clojure.core.matrix. Neanderthal implementation can be enabled.

Cartpole

$ clj -M -m examples.cartpole    ;; classic cartpole