What python module to use for neural net application?

Question

I am looking for an easy to use python module I can use to create and train a neuronal network for pattern recognition (or, more precise, I have several short audio files containing a spoken word or a spoken phrase, and I want to use a machine learning/deep learning algorithm to train the network so it can recognize which word/phrase was spoken).

So basically I have several input data elements (i.e. the same word spoken from the same person 10 times, 20 timex etc), and I have the output word/phrase.

What is a good python module to use for this kind of training?

Answer 1

Keras with the Tensorflow backend:

• Docs
• MNIST Tutorial

Installation

As a beginner, I recommend NOT to use a GPU. Things will be way slower, but you will have an easier time with the installation. You can always uninstall and install CUDA + Tensorflow with GPU support later.

1. Install Tensorflow
2. Install Kears:

It could be as simple as that:

$ pip install tensorflow
$ pip install keras

Answer 2

If you are a beginner, you may like to use EpyNN.

EpyNN is an educational ecosystem that provides:

• Educational API
• Educational workflows
• Extensive documentation with code/maths/diagrams/text

EpyNN was validated against Tensorflow/Keras and it provides identical results for identical configurations.

If you want to understand what's behind the API you use, give a try to EpyNN.