java – Performance Encog vs Deeplearning4J

We are developing Java projects using neural networks We want to test different network structures on our data set Now let's evaluate which Java neural network is the best in performance We are evaluating encog, neuroph and dl4j Can you tell us some good resources or your own experience?

Solution

Deeplearning 4J founder here:

>Encog was written by Jeff Heaton in the early 1990s and has long been a standard Java DL framework I don't believe encog handles distributed computing and can be used with GPU, Hadoop, spark or Kafka, or considering many algorithm advances in DL since 2006 (Jeff, if I'm wrong, please correct me!) > Deep learning 4J does all these things It is applicable to distributed CPUs or GPUs that use spark as the access layer It has been certified on cdh5 and will soon be certified on HDP... It includes the implementation of LSTM (RNN), deep convolution network, RBM, DBN, word2vec and other neural networks It is the most popular JVM DL tool and one of the top 5 DL libraries in the world. > Deep learning 4J is supported by numerical calculation lib nd4j or n-dimensional array of Java Basically, we ported numpy to the JVM This makes dl4j scalable, and you will see us add other algorithms, such as reinforcement learning, in the near future Conversely, nd4j runs on libnd4j, a fast computing C library We also construct a vectorization library Canova, which can obtain any type of data and convert it into vectors that can be understood by neural networks We are trying to solve some ETL problems upstream of NN. > Neuroph has strong visibility, but I can't judge the rest of the framework, so I'll let them speak for themselves!

http://deeplearning4j.org

https://github.com/deeplearning4j

https://github.com/deeplearning4j/nd4j

https://github.com/deeplearning4j/libnd4j

On gitter, the user support channel of deep learning 4J has nearly 2000 developers If you have any questions, please join us:

https://gitter.im/deeplearning4j/deeplearning4j