METHODS AND SYSTEM FOR PROVIDING SIMULTANEOUS MULTI-TASK ENSEMBLE LEARNING

• US 20150269495A1
• Filed: 06/08/2015
• Published: 09/24/2015
• Est. Priority Date: 03/13/2013
• Status: Abandoned Application

• Current Assignee: Dstillery, Inc.
• Original Assignee: Dstillery, Inc.
• Inventors: Dalessandro, Brian, Hook, Rodney, Manhendran, Yuva
• Application Number:
  US14/733,299
• Publication Number:
  US 20150269495A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/20 Ensemble learning
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

A complete end-to-end modeling system is provided that includes data sampling, feature engineering, action labeling, and model learning or learning from models built based on collected data. The end-to-end modeling process is performed via an automatic mechanism with minimal or reduced human intervention. A processor-readable medium is disclosed, storing processor-executable instructions to instantiate an automated data sampling and prediction structure training component, the automated data sampling and prediction structure training component being configured to automatically collect user event data samples, and use the collected user event data samples to train multiple prediction structures in parallel.

Claims

1. A system, comprising:

a user data sampling component configured to receive log data associated with a plurality of events, each event from the plurality of user events associated with a user from a plurality of users;

a plurality of feature engineering scheme modules, each feature engineering scheme module from the plurality of feature engineering scheme modules configured to extract features from the log data received by the user sampling component, the plurality of feature engineering scheme modules configured to extract features in parallel to simultaneously generate a plurality of training data sets; and

a machine learning platform configured to build a plurality of prediction structures, the plurality of prediction structures configured to be trained in parallel with the plurality of training data sets.