Pharo-AI - Machine Learning in Pharo

• Pharo-AI - Machine Learning in Pharo
  • Description
  • Installation
  • Implemented Algorithms
    • Machine Learning
    • Data Mining
    • Natural Language Processing
    • Data Preprocessing
    • Datasets
    • Metrics
    • Applications
    • Tools
    • Work in Progress
  • Pharo-AI and Pharo-Launcher

Description

A meta package to load all related Machine Learning libraries in Pharo.

For more information about the naming convention or the requirement to be integrated in Pharo-AI, refere to the Contribution file.

Installation

You can load all the Artificial Intelligence packages from the pharo-ai project into a fresh Pharo image by going to the Playground (Ctrl + OW/Cmd + OW) and executing the following expression (select it and press Do-it button or Ctrl+D/Cmd+D):

EpMonitor disableDuring: [
    Metacello new
        baseline: 'AIPharo';
        repository: 'github://pharo-ai/ai/src';
	onWarningLog;
	onConflictUseIncoming;
        load ]

This should load the default version of the AIPharo (you can also specify another version or branch).

To add it to your Baseline:

    spec
	    baseline: 'AIPharo'
	    with: [ spec repository: 'github://pharo-ai/ai/src' ]

If you are new to baselines and Metacello, check out the Baselines tutorial on Pharo Wiki.

Implemented Algorithms

Machine Learning

Name	Chapter	"How to use"	Coverage	CI Status	Dependencies
Linear Regression	WiP	Yes			None
Logistic Regression	No	Yes			None
K-Means	No	No			AIEditDistances
K-Nearest Neighbours	No	Yes			AIEditDistances
Naive Bayes Classifier	No	No			None
Decision Tree Model	No	Yes			DataFrame, AIDatasets

Data Mining

Name	Chapter	"How to use"	Coverage	CI Status	Dependencies
A-Priori	Yes	Yes			ContainersOrderedSet

Natural Language Processing

Name	Chapter	"How to use"	Coverage	CI Status	Dependencies
N-gram Model	No	Yes*
Term Frequency - Inverse Document Frequency	Yes	Yes			None

Data Preprocessing

Name	Chapter	"How to use"	Coverage	CI Status	Dependencies
Data Partitioners	No	Yes			None
Data Imputers	No	Yes			None
Data Preprocessing	No	Yes			None

Datasets

Name	"How to use"	Coverage	CI Status	Dependencies
Datasets	Yes			External Dependencies (DataFrame)

Metrics

Name	Chapter	"How to use"	Coverage	CI Status	Dependencies
Metrics	No	Yes			External Dependencies (VectorMatrix)
Edit Distances	No	Yes			None

Applications

Name	"How to use"	Coverage	CI Status	Dependencies
Spelling Correction	Yes			None

Tools

Name	"How to use"	Coverage	CI Status	Dependencies
Data inspector	Yes			External Dependencies (DataFrame, Roassal3)

Work in Progress

Machine Learning: 

Name	"How to use"	Coverage	CI Status	Dependencies
Gaussian Mixture Model	No	No
Hierarchical Clustering	No	No

Pharo-AI and Pharo-Launcher

Pharo-launcher is a great tool to manage Pharo images and here we are going to explain how to be able to get pharo-ai images from it.

Pharo launcher comes with a default set of sources to fetch Pharo images. It also allows one to add its own sources, and we will show here how to add Pharo-AI as one of your own source.

It is doable thanks to the mysources.list of pharo-launcher file located in the PhLUserTemplateSources sourcesFile folder. If present, this file defines additional template sources beyond the official list of templates. At this time, there is no UI to add them.

To find the right folder:

• Open the Pharo Launcher
• Open a Playground (Ctrl + O, Ctrl + W)
• Execute PhLUserTemplateSources sourcesFile

You can now edit mysources.list in this folder to add the Pharo-AI images you wish to have in your Pharo launcher. Here is an example:

[
	PhLTemplateSource {
		#type : #URLGroup,
		#name : 'Pharo-AI',
		#templates : [
			PhLTemplateSource {
				#type : #URL,
				#name : 'AI Pharo 10 - master',
				#url : 'https://github.com/pharo-ai/ai/releases/download/continuous/Pharo-AI-Pharo64-10.zip'
			},
			PhLTemplateSource {
				#type : #URL,
				#name : 'AI Pharo 11 - master',
				#url : 'https://github.com/pharo-ai/ai/releases/download/continuous/Pharo-AI-Pharo64-11.zip'
			},
			PhLTemplateSource {
				#type : #URL,
				#name : 'AI Pharo 10 - v0.8.0',
				#url : 'https://github.com/pharo-ai/ai/releases/download/v0.8.0/Pharo-AI-Pharo64-10.zip'
			},
			PhLTemplateSource {
				#type : #URL,
				#name : 'AI Pharo 11 - v0.8.0',
				#url : 'https://github.com/pharo-ai/ai/releases/download/v0.8.0/Pharo-AI-Pharo64-11.zip'
			}
		]
	}
]

This templates will add a Pharo-AI group in the templates of the Pharo-Launcher. This group will contain 4 entries:

• Pharo AI built from the latest commit of master in Pharo 10
• Pharo AI built from the latest commit of master in Pharo 11
• Pharo AI built from release v0.8.0 in Pharo 10
• Pharo AI built from release v0.8.0 in Pharo 11

You can then adapt those sources to what you need