e application of deep learning to search ranking was one of the most impactful product improvements at Airbnb. But what comes next a er you launch a deep learning model? In this paper we describe the journey beyond, discussing what we refer to as the ABCs of improving search: A for architecture, B for bias and C for cold start. For architecture, we describe a new ranking neural network, focusing on the process that evolved our existing DNN beyond a fully connected two layer network. On handling positional bias in ranking, we describe a novel approach that led to one of the most signi cant improvements in tackling inventory that the DNN historically found challenging. To solve cold start, we describe our perspective on the problem and changes we made to improve the treatment of new listings on the platform. We hope ranking teams transitioning to deep learning will nd this a practical case study of how to iterate on DNNs.
Craniosynostosis is the premature fusion of the bones of the calvaria resulting in abnormal skull shapes that can be associated with increased intracranial pressure. The goal of this work is to analyze the various 3D skull shapes that manifest in isolated single suture craniosynostosis. A logistic regression is used to identify different types of synostosis and quantify the differences. Due to the high-dimensionality of the feature data, a sophisticated feature selection technique is required to avoid overfitting and to improve the classification accuracy on the unseen data. In addition, feature selection allows the identification of surface areas that contribute to the major skull deformations that characterize isolated synostosis. We applied three sparse feature selection methods: L1 regularization (lasso [9]), fused lasso ([10]) and a novel regularization method we have developed called the clustering lasso (cLasso). L1 regularized logistic regression locates important surface points, and the fused lasso groups these points into regions. The cLasso was designed to assign similar weights to groups of correlated shape features. Experimental results indicated that the regularized logistic regression models achieve a significantly lower misclassification rate than unregularized logistic regression.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.