Table of contents 1. Introduction. 2. Online predictions with batch features. 3. Online predictions with online features: real time vs near real time. 4. When does it make sense to move things online? Introduction As a follow up of my past article related to Batch Serving: Machine Learning at scale:...
Table of contents 1. Introduction. 2. What is a Feature Store anyway? 3. Offline Batch Layer 4. Online Serving Layer 5. Closing thoughts Introduction In today's article I will discuss engineering challenges faced by Constructor.io [1] when implementing a feature store. Let's dive in! What is a Feature Store...
Table of contents 1. Introduction. 2. Adversarial examples and robust classifiers. 3. How to generate adversarial examples. 4. How to defend your precious Machine Learning models against adversarial examples. Introduction Today's article will dive deep into adversarial examples. There are two major reasons adversarial examples are important to understand: 1....
Table of contents 1. Introduction. 2. Past actions affect personalized recommendations with delay. 3. Leveraging actions in near real-time to adapt recommendations. 4. Results. Introduction In today's article, I will talk about LinkedIn transitioned from an offline update to a near real-time update of features for recommender systems that power...
Table of contents 1. Introduction. 2. Trading off explainability with other metrics. 3. Inherently explainable models. 4. Improving explainability in non-inherently explainable models. Introduction In this article, I will talk about the need of having explainable and interpretable models. Explainability and interpretability are often used interchangeably, but I'd like to...
Table of contents 1. Introduction. 2. First iteration: on-demand batch processing. 3. Second iteration: enabling online requests with pre-computation. 4. Final system architecture. Introduction In today's article, I will describe how Netflix built a ML Media understanding platform. Let's start from the use cases: 1️⃣Dialogue search An editor should...
