I'm Tatsuki Koga, a 3rd-year Ph.D. student in Computer Science at University of California San Diego, advised by Prof. Kamalika Chaudhuri.

My research interests are machine learning with privacy guarantees, under a class imbalance, and for medical applications.

I have obtained my B.S. in Information Science at the University of Tokyo in March 2020.

CV

Blog

Talk & Slides

Contact

Education

• 2020.9 - Present (Expected Graduation 2025): Ph.D. in Computer Science, University of California San Diego
• 2016.4 - 2020.3: B.S. in Information Science, The University of Tokyo
  • 2018.4 - 2020.3: School of Science, Dept. of Information Science
  • 2016.4 - 2018.3: College of Arts and Sciences
• 2013.4 - 2016.3: Tokyo Gakugei University Senior High School

Publications

• Koga, T., Song, C., Pelikan, M., Chitnis, M. ”Population Expansion for Training Language Models with Private Federated Learning”. Federated Learning and Analytics in Practice: Algorithms, Systems, Applications, and Opportunities Workshop at 40th International Conference on Machine Learning (ICML), Hawaii, United States, July 2023.
  • [Paper]
• Koga, T., Meehan, C. & Chaudhuri, K. "Privacy Amplification by Subsampling in Time Domain". 25th International Conference on Artificial Intelligence and Statistics (AISTATS), Virtual, Mar. 2022.
  • [Paper]
  • A short version has been presented at TPDP 2021 - Theory and Practice of Differential Privacy at 38th International Conference on Machine Learning (ICML), Virtual, Jul. 2021.
• Koga, T., Rie, E., Hirose, K. & Seita, J. “Human and GAN collaboration to create haute couture dress”. NeurIPS Workshop on Machine Learning for Creativity and Design 3.0 at 33rd Conference on Neural Information Processing Systems (NeurIPS), Vancouver, Canada, Dec. 2019. Contributed Talk.
  • [Video]
• Koga, T., Nonaka, N., Sakuma, J. & Seita, J. “General-to-Detailed GAN for Infrequent Class Medical Images”. ML4H: Machine Learning for Health Workshop at 32nd Conference on Neural Information Processing Systems (NeurIPS), Montreal, Canada, Dec. 2018.
  • [Paper]

Preprints

• Koga, T., Chaudhuri, K. & Page, D. ”Differentially Private Multi-Site Treatment Effect Estimation”.
  • [Paper]

Exhibitions

• TOKYO Fashion Week 2019 A/W Collection by EMarie, February 2019
  • [Link], [Video]

Experiences

• Microsoft Corporation, Data Scientist Intern, Summer 2023
  • Worked at Data Security and Privacy Research team and worked on applied research on compliance.
  • Developed a novel pipeline for Microsoft Purview Data Loss Prevention solution using the embedding learned from scratch with neural networks.
• Apple Inc., ML Engineer Intern, Summer 2022
  • Worked at Private Federated Learning (PFL) team and worked on building language models (LM) in PFL.
  • Conducted research on domain adaptation technique for LM in PFL to deal with the setting where we have smaller population size for specific use-cases.
• RIKEN Medical Science Innovation Hub Program (MIH), AI based Healthcare and Medical Data Analysis Standardization Unit, Research Part-time Worker II, Fall 2017 - Summer 2020
  • Conducted research on generating medical data (clinical data, images) fed into machine learning methods with GANs in order to resolve the imbalance of the classes in datasets.
• RIKEN Advanced Intelligence Project (AIP), AI Security and Privacy Team, Student Trainee, Spring 2018 - Summer 2020
  • Conducted research on differentially private deep neural network, especially generative models for medical data and its tradeoff between model’s utility and privacy.