About

About Lab

Our research interest lies in the intersection of
data mining, machine learning, and optimization.

In the perspective of developing methodologies,
it focuses on investigating the source of possible leverage in improving data mining and machine learning methods from real-world applications.
Applying machine learning and optimization techniques to the real-world problems is another direction of our research.

Since, most of the times, the design of new algorithms is required by a specific application, and doing so provides strategic advantages,
we have been doing our research to perform the former and latter simultaneously.

We also have an interest in applying data-driven models to system control for implementing manufacturing intelligence and other system intelligence.
Working with various types and large amount of data from real fields,
the overall goal of our research is to help organizations make better decisions based on scientific evidences and to create intelligence systems.

Research Areas

Methodologies

01Physics-Informed Learning

  • Incorporating prior knowledge or physics such as monotonicity into data-driven models (Monotonic Neural Networks, Physics-Informed Neural Networks, etc.)

02Sustainable Learning / Continual Learning

  • Monitoring changes in data distribution and incorporates them into model retraining decisions to ensure long-term validity of predictive models
    (Temporal Peak-Management and Clustering System, Domain Adaptation, etc.)

03Interpretable Machine Learning

  • Enabling interpretability of data-driven (AI) models by explaining their operational principles and inferences
    (Mixture of Local Interpretable Experts, Prediction Reliability, etc.)

04Learning from Tabular (Time-Series) Data

  • Developing new learning techniques for tabular time-series data by adapting and refining existing deep learning methodologies primarily designed for
    image or text data

05System Control and Optimization

  • The results from the above research works have the potential to revolutionize the field of system control & optimization, enhancing its efficiency and effectiveness. (DLO (Data/Learning/Optimization) Approach, Bayesian Optimization)

Applications

01Manufacturing Intelligence

  • Process control and real-time decision-making
  • Equipment performance optimization
  • Waste reduction and resource optimization
  • Yield optimization
  • Energy management
  • Product design and customization
  • Quality control
  • Predictive analytics for maintenance
  • Defect detection and classification
  • Anomaly detection

02Service Intelligence

  • Service process optimization
  • Resource allocation and scheduling
  • Energy management and efficiency
  • Predictive maintenance in service equipment
  • Price forecasting for fleet management
  • Financial fraud detection