About the Research
Reaction network modeling, data-driven prediction, optimal design of experiments
Machine learning is about developing computer algorithms that can detect patterns in data without being explicitly programmed for any specific pattern. Usually, the process is designed for tabular data, but much scientific data is not in this form. For example, genomes are sequential data, and structural formulas of chemical compounds are network-like graphical data. My special focus is to develop machine learning algorithms that can handle these kinds of non-numerical data.
The ability to process the various kinds of data generated by chemical experiments and simulations is indispensable for rationally designing chemical reactions. With cutting-edge machine learning, I hope to make full use of data and theory to uncover the highly complex nature of real-world chemical reactions. This can contribute to modelling uncertain factors, predicting any promising targets and conditions, extracting new knowledge on determining factors, and seamlessly integrating theory-driven, knowledge-driven, and data-driven information.
The Researcher’s Perspective
Machine learning is a kind of meta-science due to its applicability to many different scientific problems, and there seem to be deep relationships to how we humans learn, ourselves. However, it is very difficult to explain the human learning process explicitly, and so it is a challenging but interesting task to teach a machine to learn or to discover something new.
Representative Research Achievements
- Toward effective utilization of methane: machine learning prediction of adsorption energies on metal alloys
Toyao T, Suzuki K, Kikuchi S, Takakusagi S, Shimizu K, Takigawa I. J. Phys. Chem. C, 2018, 122 (15): 8315-8326
DOI : 10.1021/acs.jpcc.7b12670
- Generalized sparse learning of linear models over the complete subgraph feature set
I Takigawa, H Mamitsuka, IEEE Trans. Pattern Anal. Mach. Intell. 2017, 39 (3): 617-624
DOI : 10.1109/TPAMI.2016.2567399
- Machine-learning prediction of d-band center for metals and bimetals
I Takigawa, K Shimizu, K Tsuda, S Takakusagi, RSC Adv. 2016, 6: 52587-52595
DOI : 10.1039/C6RA04345C
- MED26 regulates the transcription of snRNA genes through the recruitment of little elongation complex
H Takahashi, I Takigawa, M Watanabe, D Anwar, M Shibata, C Tomomori-Sato, S Sato, A Ranjan, C W Seidel, T Tsukiyama, W Mizushima, M Hayashi, Y Ohkawa, J W Conaway, R C Conaway, S Hatakeyama, Nat. Commun. 2015, 6 (5941)
DOI : 10.1038/ncomms6941
- Efficiently mining δ-tolerance closed frequent subgraphs
I Takigawa, H Mamitsuka, Machine Learning, 2011, 82 (2): 95-121
DOI : 10.1007/s10994-010-5215-6