nd4clj

An implementation of core.matrix protocols with nd4j. Most importantly this is supposed to allow the usage of the jcublas backend of nd4j and the integration of deeplearning4j code in Clojure projects.

Requirements

• be core.matrix compliant
• don't introduce significant performance overhead
• make implementation of machine learning algorithms in Clojure more feasible
• avoid wrapping INDArray Matrices for direct interop with deeplearning4j

TODO

• make shaping of Nd4j compatible with core.matrix, e.g. introduce Vector type, see https://github.com/mikera/core.matrix/wiki/Vectors-vs.-matrices
• fix other outstanding issues in compliance tests
• evaluate the GPU backend in comparison to theano, e.g. with boltzmann
• implement example with deeplearning4j and core.matrix dependent code, e.g. boltzmann, incanter 2.0

Usage

Not really usable yet, as it is not compliant to core.matrix.

License

Copyright © 2015 Christian Weilbach

Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.