2023 Fiscal Year Final Research Report
Development of statistical analysis methods for visualizing nonlinear activity of large-scale neural populations
| Project/Area Number |
20K11709
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 60030:Statistical science-related
|
|---|
| Research Institution | Kyoto University (2022-2023)
Hokkaido University (2020-2021) |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2020-04-01 – 2024-03-31
|
| Keywords | 神経スパイクデータ / イジングモデル / 状態空間モデル |
|---|
| Outline of Final Research Achievements |
We studied the nonequilibrium kinetic Ising model to analyze and visualize large-scale, nonlinear neuronal population activity. We developed a theory that unifies various mean-field approximation methods and proposed a new method that outperforms previous ones in conditions where neurons exhibit diverse patterns. We also conducted theoretical analyses on this model using a path-integral approach to elucidate the behavior of large-scale networks. Furthermore, we extended the nonequilibrium kinetic Ising model within the state-space framework. This model enabled us to estimate fluctuating asymmetric couplings from experimentally obtained spiking time-series data. In conclusion, we successfully developed novel theories and statistical analysis methods for understanding and visualizing large-scale neuronal population activity.
|
| Free Research Field |
理論神経科学
|
| Academic Significance and Societal Importance of the Research Achievements |
非平衡キネティック・イジングモデルに対する新しい理論と統計解析技術は，神経科学分野における基礎研究だけでなく，機械学習分野にも応用可能である．提案した平均場近似法によって機械学習による大規模データ解析の精度が向上が見込まれる．さらに，提案した平均場近似の統一的枠組みに基づいて，機械学習モデルのトランスフォーマーの解析が海外の研究者によって行われるなど，学習機械の理論的な理解にも一定の貢献がある．また，理論解析により求めた大規模ネットワークの厳密解は他の複雑系の解析にも適用できるため，物理学，生物学，経済学など多岐にわたる分野で観測されるネットワーク現象への応用が期待される．
|
