Bio
Inbar is a Data Scientist at Taboola, where she applies deep learning techniques for content recommendations. In the past, she worked with different types of data, including DNA sequences, neurological recordings, click streams, texts and images.
She has an MS.c in Computer Science with a focus on machine learning research, and a BS.c in Computer Science and Cognitive Science. In her spare time she is the host of Unsupervised – a podcast about data science in Israel; a co-founder and manager of DataHack, a Data Science and Machine Learning Hackathon and the DataTalk meetup.
Bio
Inbar is a Data Scientist at Taboola, where she applies deep learning techniques for content recommendations. In the past, she worked with different types of data, including DNA sequences, neurological recordings, click streams, texts and images.
She has an MS.c in Computer Science with a focus on machine learning research, and a BS.c in Computer Science and Cognitive Science. In her spare time she is the host of Unsupervised – a podcast about data science in Israel; a co-founder and manager of DataHack, a Data Science and Machine Learning Hackathon and the DataTalk meetup.
Abstract
Deep Learning models have been gaining increasing attention in the recommendation systems community, replacing some of the traditional methods. The sparse nature of the problems in this domain and the different types of inputs offer unique challenges for feature engineering and architecture planning, in order to balance between memorization and generalization.
During the past two years the algorithms team in Taboola moved all of our algorithms to deep learning models and in this talk we will share the lessons we learned doing so. Specifically, we will talk about building neural networks with multiple input types (click history, text and pictures); feature engineering in the deep learning era; embeddings for categorical features; capturing interactions between features using both deep dense architectures and Factorization Machine models; Tradeoffs between deep models, shallow models and the combination of the two; and other tips regarding network architectures.
Abstract
Deep Learning models have been gaining increasing attention in the recommendation systems community, replacing some of the traditional methods. The sparse nature of the problems in this domain and the different types of inputs offer unique challenges for feature engineering and architecture planning, in order to balance between memorization and generalization.
During the past two years the algorithms team in Taboola moved all of our algorithms to deep learning models and in this talk we will share the lessons we learned doing so. Specifically, we will talk about building neural networks with multiple input types (click history, text and pictures); feature engineering in the deep learning era; embeddings for categorical features; capturing interactions between features using both deep dense architectures and Factorization Machine models; Tradeoffs between deep models, shallow models and the combination of the two; and other tips regarding network architectures.
