Data Science / AutoML

PFN’s data science discovers valuable knowledge by leveraging our large-scale computing resources and real-world data. We research and develop machine learning technologies in a variety of research fields, such as time series prediction, anomaly detection, mathematical optimization, system control, and physical simulation. We are also working on machine learning applications to solve challenging real-world problems in various industries including the semiconductor, energy, automobile, production facilities, and retail industries.
These R&D involves much trial and error in processes such as feature engineering, model selection, architecture search, and hyperparameter optimization. We also research automated machine learning technologies and develop Optuna, a hyperparameter optimization framework. We streamline our workflow and make our optimization optimized.