Search

An interview with Naonori Ueda, who is the director of Machine Learning at the Data Service Center and Senior Distinguished scientist at Nippon Telegraph and Telephone Communication Science Laboratories, is presented. He talks about his career, the definition of machine learning, and the challenges in big data analysis.

We introduce technology for predicting the risk of congestion in the near future from real-time observational data of people or automobile flows and for automatically deriving an optimal crowd navigation plan online to avoid that risk. This technology can deal with unforeseen situations by performing simulations based on machine learning. Our aim in establishing this technology is to help construct a safe and secure and more comfortable social infrastructure. [ABSTRACT FROM AUTHOR]

This paper presents the Averaged CVB (ACVB) inference and offers convergence-guaranteed and practically useful fast Collapsed Variational Bayes (CVB) inferences. CVB inferences yield more precise inferences of Bayesian probabilistic models than Variational Bayes (VB) inferences. However, their convergence aspect is fairly unknown and has not been scrutinized. To make CVB more useful, we study their convergence behaviors in a empirical and practical approach. We develop a convergence-guaranteed algorithm for any CVB-based inference called ACVB, which enables automatic convergence detection and frees non-expert practitioners from the difficult and costly manual monitoring of inference processes. In experiments, ACVB inferences are comparable to or better than those of existing inference methods and deterministic, fast, and provide easier convergence detection. These features are especially convenient for practitioners who want precise Bayesian inference with assured convergence. [ABSTRACT FROM AUTHOR]

Second-order polynomial regression can often outperform simple linear regression by making use of feature combinations. However, when the number of feature combinations is large, second-order polynomial regression quickly becomes impractical. In this article, we present convex factorization machines, a new technology developed by NTT Communication Science Laboratories, which can cope with a large number of feature combinations and guarantees globally optimal model parameters. [ABSTRACT FROM AUTHOR]

We present an application of machine-learning (ML) techniques to source selection in the optical transient survey data with the Hyper Suprime-Cam (HSC) on the Subaru telescope. Our goal is to select real transient events accurately and in a timely manner out of a large number of false candidates, obtained by the standard difference-imaging method. We have developed the transient selector, which is based on majority voting of the three ML machines of AUC Boosting, Random Forest, and Deep Neural Networks. We applied it to our observing runs of Subaru-HSC in 2015 May and August, and proved it to be efficient in selecting optical transients. The false positive rate was 1.0% at the true positive rate of 90% in the magnitude range of 22.0-25.0mag for the May data. For the August run, we successfully detected and reported 10 supernovae candidates within the same day as the observation. From these runs, we learned the following lessons: (1) training using artificial objects is effective in filtering false candidates out, especially for faint objects, and (2) a combination of ML by majority voting is advantageous. [ABSTRACT FROM AUTHOR]