Metadata-Version: 2.1
Name: hula
Version: 0.0.0.2
Summary: Experimental learning methods. The first appearance of these methods.
Home-page: https://github.com/robertbean1/Hula
Author: Robert Bassett
Author-email: robert.bassett.coder@gmail.com
License: UNKNOWN
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: GNU General Public License (GPL)
Classifier: Operating System :: OS Independent
Description-Content-Type: text/markdown
Requires-Dist: numpy

# hula-0.0.0.2

## What is hula?

Hula is a set of unconventional machine learning modules. Hula is designed to unify the desired aspects of most machine learning capabilities out there. 

Within hula are a few machine learning algorithms which I've coined as the following:

-  Deep Recursive Learning
-  Reinforcement via Similarity
-  Comprehensive Learning

Respectively, the modules are named:

- RecursiveL
- ReinforcementL
- ComprehensiveNet

I will be working on a blog to explain the workings behind each algorithm.

## Documentation

---

### hula.ComprehensiveNet.CNET(*design*)

Generates an Artificial Neural Network comprised of Memory Activation nodes corresponding to the dimensions of *design*

---

#### hula.ComprehensiveNet.CNET.activate(*X*)

Feeds X through the network and returns the output of the last layer.

---

#### hula.ComprehensiveNet.CNET.act(*alpha*)

Generates a limbo-action proportional to alpha for each memory node in the network.

---

#### hula.ComprehensiveNet.CNET.score(*score*)

Retroactively scores each limbo-action and turns them into state-actions. Higher scores are favored.

---

#### hula.ComprehensiveNet.CNET.train(*alpha*)

Finds the lowest distance to each action's state minus the score of that action.

---

#### hula.ComprehensiveNet.CNET.trim(*perc*)

removes perc percent of state-actions from the state-action tree

----


This list is not complete. There are many more modules, classes, and methods to cover.

