Pre-Release

Changelog

Verbosity, documentation and UI updates

---

See the documentation at https://epistasislab.github.io/pennai/ for more instructions.

Requirements:

See the installation instructions for prerequisite software requirements.

Installation:

1. Download the production zip pennai-0_13.zip from the Assets section below (be sure not to download the source code zip).
2. Unzip the archive

Running:

See Using PennAI for instructions.

• From the pennai directory, run the command docker-compose up to start the PennAI server.
• To stop PennAI, kill the process with ctrl+c and wait for the server to shut down.
• Once the webserver is up, connect to http://localhost:5080/ to access the website.

Pre-release

Pre-release of PennAI.

Includes svd, knn, average, and random recommenders.

Installation:

1. Download pennai-0_12.zip and unzip
2. From the command line, navigate to the pennai directory and load the images into docker with the following commands:

docker load --input .\images\pennai_lab.tar
docker load --input .\images\pennai_machine.tar
docker load --input .\images\pennai_dbmongo.tar

Running:

From the pennai directory, run the command docker-compose up to start the PennAI server. See the quickstart for more instructions.

Feature Updates

• User datasets can now be added by putting .csv or .tsv files in the data/datasets/users folder
• KNN recommender implemented
• Recommenders can be bootstrapped with a knowledgebase of previous results
• Usability updates to the fitted model and example script that can be downloaded from the Experiments page for completed experiments
• Users can specify the 'target' column for a dataset
• Sphinx developer documentation added for the ai and recommender python code
• Major refactoring of the ai engine
• Major refactoring of ml code in machine instances
• Major dataset handling refactoring
• Docker test environment and test runner added for unit tests
• Jenkins CI configuration added to run unit and integration tests and build documentation

Feature updates

• Dataset metafeatures are generated and stored during dataset upload.
• Supports random, meta and average recommender. Need to add knowledgebase for meta recommender to be useful.
• Added UI datasets details page. Shows preview of the data, metafeatures.
• UI improvements. Algo and parameter descriptions and documentation links, 'no capacity' errors if no machines available when attempting to start an algorithm, etc.
• Improved build process. node_modules generated during docker build instead of local build, smaller images, removed some unnecessary packages, etc.
• SemanticUI CSS added to project, so once it's been built internet access is no longer required to run.
• Python unit tests with code coverage.

v0.9: error message bug fix

Stable pre-release. Supports random recommender, single machine.

v.0.1: Brief debug log

Former-commit-id: 1db265ffbf32da9651b1ca69e5a44cbea0a61f0d