Tensorflow with Keras for Neural Prediction

Project Owner

  • Jim Cook (ICT, Techlab)

Project Credits

  • Jim Cook (ICT, Techlab)

Start Date

18th December 2017

Status

Initial Training and demo completed 22nd December 2017

Final training epoch completed 17th january 2017

Problem Statement

One of a series of Neural Network experiments in various languages, to understand the core types of Machine learning that are becoming prevalent in industry and academia. This example leverages the Keras package for Python as it is a rapid prototyping environment for extensibility. This example is based on the concept of Long short-term memory.

Final Brief

We developed a system to provide an overall assessment of the various Keras regressions. While We don’t claim to be expert, it has proved easy to side by side compare these with a good starting data set. The question this code actually poses is, could you learn to write Progressive metal songs if all you had was the lyrics to all of Ayreon, Symphony X, Seventh Wonder, and Pagans Mind. More importantly could you learn to write ENGLISH with just these songs? This includes Grammar, Nouns, verbs and even stage directions.

Challenges & Learnings

  • C1. Python on Windows.
  • L1. This is essentially a problem that we will learn to live with. In the future we should use frameworks like Conda or Anaconda, to avoid dependency and installation issues.
  • C2. Not freaking out when Arjen Anthony Lucassen, the Mind behind Ayreon and Star One, comments on your facebook post about this.

Languages / Framework

Python, Keras, Tensorflow, CUDA, HDF5

Links to Resources

Github Repo



Tags: