Distributed LSTM-Learning from Differentially Private Label Proportions

TL;DR

This paper introduces two decentralized LSTM models that incorporate Differential Privacy to learn from label proportions in spatio-temporal data, balancing privacy and performance.

Contribution

It proposes novel models combining Differential Privacy with decentralized LSTM learning for label proportion data, extending existing methods with histogram-based neighbor information integration.

Findings

Effective privacy-performance tradeoff demonstrated on multiple datasets

Models successfully learn from label proportions while preserving privacy

Evaluation shows competitive accuracy with privacy guarantees

Abstract

Data privacy and decentralised data collection has become more and more popular in recent years. In order to solve issues with privacy, communication bandwidth and learning from spatio-temporal data, we will propose two efficient models which use Differential Privacy and decentralized LSTM-Learning: One, in which a Long Short Term Memory (LSTM) model is learned for extracting local temporal node constraints and feeding them into a Dense-Layer (LabelProportionToLocal). The other approach extends the first one by fetching histogram data from the neighbors and joining the information with the LSTM output (LabelProportionToDense). For evaluation two popular datasets are used: Pems-Bay and METR-LA. Additionally, we provide an own dataset, which is based on LuST. The evaluation will show the tradeoff between performance and data privacy.