Analysis to gain new knowledge from huge amounts of data is called data science, and its widespread use is now socially important. Feature engineering, the process of extracting features from data, is one of the main tasks in data science, and since this task relies on the experience of experts, research is being conducted to automate it. In this paper, we propose a novel approach to automate feature identification from textual information in data column names. Specifically, we use techniques of natural language processing and source code analysis, for data descriptions and source codes in Python notebooks to create a knowledge database with a particular focus on datetime features. We develop a recommendation system of datetime features for newly given text information based on that knowledge database. In experiments, we confirmed the classification accuracy of the knowledge database, applied the database to actual forecasting tasks such as home price forecasting, and achieved 5.76% on average for accuracy gain.

In recent years, the automation of machine learning and data science (AutoML) has attracted significant attention. One under-explored dimension of AutoML is being able to automatically utilize domain knowledge (such as semantic concepts and relationships) located in historical code or literature from the problem's domain. In this paper, we demonstrate Semantic Feature Discovery, which enables users to interactively explore features semantically discovered from existing data science code and external knowledge. It does so by detecting semantic concepts for a given dataset, and then using these concepts to determine relevant feature engineering operations from historical code and knowledge.

Autonomous agents need to make decisions in a sequential manner, under partially observable environment, and in consideration of how other agents behave. In critical situations, such decisions need to be made in real time for example to avoid collisions and recover to safe conditions. We propose a technique of tree search where a deterministic and pessimistic scenario is used after a specified depth. Because there is no branching with the deterministic scenario, the proposed technique allows us to take into account the events that can occur far ahead in the future. The effectiveness of the proposed technique is demonstrated in Pommerman, a multi-agent environment used in a NeurIPS 2018 competition, where the agents that implement the proposed technique have won the first and third places.

The anomaly localization task aims at detecting faulty sensors automatically by monitoring the sensor values. In this paper, we propose an anomaly localization algorithm with a consistency guarantee on its results. Although several algorithms were proposed in the last decade, the consistency of the localization results was not discussed in the literature. To the best of our knowledge, this is the first study that provides theoretical guarantees for the localization results. Our new approach is to formulate the task as solving the sparsest subgraph problem on a difference graph. Since this problem is NP-hard, we then use a convex quadratic programming approximation algorithm, which is guaranteed to be consistent under suitable conditions. Across the simulations on both synthetic and real world datasets, we verify that the proposed method achieves higher anomaly localization performance compared to existing methods.

We proposed a new method (;linearized method') to analyze neuroleptic ligand-receptor specific binding in a human brain using positron emission tomography (PET). We derived the linear equation to solve four rate constants, k(3), k(4), k(5), k(6) from PET data. This method does not demand a radioactivity curve in plasma as an input function to the brain, and can perform fast calculations in order to determine rate constants. We also tested the nonlinearized method including nonlinear equations which is a conventional analysis using plasma radioactivity corrected for ligand metabolites as an input function. We applied these methods to evaluate dopamine D(2) receptor specific binding of [(11)C] YM-09151-2. The value of B(max)/K(d)=k(3)/k (4) obtained by the linearized method was 5.72+/-3.1 which was consistent with the value of 5.78+/-3.4 obtained by the nonlinearized method.