Lambda Learner: Fast Incremental Learning on Data Streams

TL;DR

Lambda Learner is a novel incremental learning framework designed for fast, real-time model updates on high-velocity data streams, outperforming traditional offline models in time-sensitive scenarios.

Contribution

The paper introduces Lambda Learner, a new incremental training framework that closely estimates offline models and improves loss over stale models, with theoretical and large-scale practical validation.

Findings

Model updates outperform stale batch models in loss reduction.

Framework effectively handles high-velocity data streams in large-scale deployment.

Theoretical proof supports incremental updates improving model accuracy.

Abstract

One of the most well-established applications of machine learning is in deciding what content to show website visitors. When observation data comes from high-velocity, user-generated data streams, machine learning methods perform a balancing act between model complexity, training time, and computational costs. Furthermore, when model freshness is critical, the training of models becomes time-constrained. Parallelized batch offline training, although horizontally scalable, is often not time-considerate or cost-effective. In this paper, we propose Lambda Learner, a new framework for training models by incremental updates in response to mini-batches from data streams. We show that the resulting model of our framework closely estimates a periodically updated model trained on offline data and outperforms it when model updates are time-sensitive. We provide theoretical proof that the incremental learning updates improve the loss-function over a stale batch model. We present a large-scale deployment on the sponsored content platform for a large social network, serving hundreds of millions of users across different channels (e.g., desktop, mobile). We address challenges and complexities from both algorithms and infrastructure perspectives, and illustrate the system details for computation, storage, and streaming production of training data.