nd4clj

Rationale

If you want to use clojure with the deeplearning4j library a clojure wrapper for nd4j functions is needed. This wrapper implements the idiomatic and familiar clojure matrix interface core.matrix. Existing clojure code using the core.matrix can change its implementation to nd4clj and take advatage of the performance of nd4j and interoperability with deeplearning4j libraries. b The NDArray fundemental to deeplearning4j code is now wrapped with this release. It can be unwraped by dereferencing the public property "a" as in (.a (indarray/matrix :nd4clj [[1 2] [3 4]])).

Tutorial

Add [org.clojars.ds923y/nd4clj "0.1.1-SNAPSHOT"] to your project.clj file. Here is an example of usage.

(ns useit.core
  (:require [nd4clj.matrix :as imp]
            [clojure.core.matrix :as indarray])
  (:gen-class))

  (indarray/set-current-implementation :nd4j)

(def A (indarray/array [[0 1] [2 3]]))
(def I (indarray/array [[1 0] [0 1]]))

(defn -main
  [& args]
  (println (indarray/mmul A I)))

Limitations

This is a largely compliant implementation. The bare minimum of protocols are supported though. Most people should not have problems.