LearnTools

Stucture

• Network

  • network (class to hold network) (need to add more insightful prints)

  • layer_dense (dense layer from layer class)

  • layer_one_to_one

  • layer_dropout

  • activation_function (custom activation function class)

  • relu / softmax / sigmoid (standard activation functions)

  • layer_conv WIP

  • layer_norm WIP (or perhaps a function to normalise data)

• Learning

  • random_learning

  • random_momentumn_learning

  • random_evolution_learning WIP

  • random_stochastic_learning (batch random) WIP

  • gradient descent WIP

  • stochastic gradient descent WIP WIP

Improvements

• Fix random_learning2
• Look into making random_learning faster
  • removal of backwards pertubation
  • scale = step in normal dist in stead of step * normal dist
• Making printing nicer
• Make learning a class instead of a function
  • Loss is maintained as a attribute
  • alongisde other useful data
• Add a layer without a bias
• Add just a bias layer
• Learning function that remove least important weights

Getting started

run the command pip install learntools
for network use from learntools import Network
for learning use from learntools import Learning

Note to Self

• Setup Venv

  • install wheel, twine

• To upload package

• build dist:

  • rmdir /s /q build dist learntools.egg-info
  • python setup.py sdist bdist_wheel

• upload dist:

  • twine upload dist/*

• To Build Documentation: https://realpython.com/python-project-documentation-with-mkdocs/#:~:text=Build%20Your%20Python%20Project%20Documentation%20With%20MkDocs%201,Step%203%3A%20Write%20and%20Format%20Your%20Docstrings%20