Presented By O’Reilly and Cloudera
Make Data Work
September 11, 2018: Training & Tutorials
September 12–13, 2018: Keynotes & Sessions
New York, NY

Solving the cold start problem: Data and model aggregation using differential privacy

Chang Liu (Georgian Partners )
2:55pm–3:35pm Wednesday, 09/12/2018
Data science and machine learning
Location: 1A 08 Level: Beginner
Secondary topics:  Ethics and Privacy
Average rating: *****
(5.00, 1 rating)

Who is this presentation for?

  • Machine learning practitioners and data scientists

Prerequisite knowledge

  • Basic math skills
  • Familiarity with machine learning and gradient descent

What you'll learn

  • Understand the cold start problem
  • Learn how to solve it using differentially private data aggregation

Description

A key factor in developing high-performing machine learning models is having large enough datasets. While collecting more intracompany data is one solution, another is to aggregate data or share information from trained machine learning models across multiple companies with similar data. Chang Liu explains how Georgian Partners has been successful at transferring knowledge from existing data through differentially private data aggregation.

Chang outlines two frameworks for building differentially private aggregation approaches to enable transferring knowledge from existing models trained on other companies’ datasets to a new company with limited or no labeled data. The two approaches are based on state-of-the-art private learning algorithms: bolt-on differentially private stochastic gradient descent and private aggregation of teacher ensemble. Chang demonstrates how using differential private techniques helps enable private data aggregation and augment data utility while providing provable mathematical guarantees on privacy. The proposed methods can provide significant business value for SaaS companies, specifically as a solution for the cold start problem.

Photo of Chang Liu

Chang Liu

Georgian Partners

Chang Liu is an applied research scientist at Georgian Partners and a member of the Georgian impact team, where she draws on her in-depth knowledge of mathematical and combinatorial optimization to help Georgian’s portfolio companies. Previously, Chang was a risk analyst at Manulife Bank, where she built models to assess the bank’s risk exposure based on extensive market research, including evaluating and predicting the impact of the oil price drop to the mortgage lending risks in Alberta in 2014. Chang holds a master of applied science in operations research from the University of Toronto, where she specialized in combinatorial optimization, and a bachelor’s degree in mathematics from the University of Waterloo.