ML21

Settings

The following environment variables need to be set:

• LK_DATABASE_NAME: name of the postgreSQL database
• LK_DATABASE_USER: user of the postgreSQL database
• LK_DATABASE_PASSWORD: password of the postgreSQL database

Current state of the project

• draft of a django project
• draft of a database structure (in PostGre)

Todo

Some ideas:

• command to apply all existing scikit learn estimators to solve the problem
• view and template to compare the results of the different estimators (and details for each estimators: best parameters)

Algorithmics

• Preprocessing : standardization, missing values, etc.
• Feature engineering : feature selection, feature transformation, signal processing, dimensionality reduction, unsupervised learning
• Classification and regression : scikit-learn + other libraries (deep learning with theano or tensorflow, gradient boosting with xgboost, etc.)
• Hyper-parameter optimization : grid-search, random search, heuristics, general purpose global optimization algorithms
• Combination of multiple algorithms : in series (PCA+SVM -> output) or in parallel (SVM || random forest -> averaged prediction)
• Automatic detection of the type of data/problem : time-series regression ? image classification ? etc.

Computing

• Parallelization
• Where are the computations done : locally on the user computer ? on a dedicated server ? on the cloud ?
• Do we need funding ? Public or private ?