Distributed labeling for supervised learning

Abstract

Embodiments described herein provide a technique to crowdsource labeling of training data for a machine learning model while maintaining the privacy of the data provided by crowdsourcing participants.Client devices can be used to generate proposed labels for a unit of data to be used in a training dataset. One or more privacy mechanisms are used to protect user data when transmitting the data toa server. The server can aggregate the proposed labels and use the most frequently proposed labels for an element as the label for the element when generating training data for the machine learning model. The machine learning model is then trained using the crowdsourced labels to improve the accuracy of the model.

Description

[0001] cross reference [0002] The present application claims priority to U.S. Patent Application No. 16 / 566066, filed on August 29, 2018, is submitted by the US Temporary Patent Application No. 62 / 73899, which is submitted in this two patent applications. Enter this article. Technical field [0003] The entire disclosure involves the field of machine learning via privatization data. More specifically, the present disclosure relates to a system that implements one or more mechanisms to enable monitoring training of machine learning models. Background technique [0004] Machine learning is an artificial intelligence application that allows complex systems to automatically learn and improve from experience without clarifying. The accuracy and effectiveness of the machine learning model can partly depend on data used to train those models. For example, a data set labeled training machine learning classifier, the classifier to learn where the sample identification data identifying th...

AI Technical Summary

Problems solved by technology

However, current techniques for preparing training datasets can be cumbersome, time-consuming, and expensive, especially those that involve manually labeling data to generate training datasets

Images