| Project/Area Number |
17K00338
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Soft computing
|
|---|
| Research Institution | Kyoto University (2018-2019)
Osaka University (2017) |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
松尾 直毅 大阪大学, 医学系研究科, 准教授 (10508956)
|
|---|
| Project Period (FY) |
2017-04-01 – 2020-03-31
|
| Project Status |
Completed (Fiscal Year 2019)
|
| Budget Amount *help |
¥4,550,000 (Direct Cost: ¥3,500,000、Indirect Cost: ¥1,050,000)
Fiscal Year 2019: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2018: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2017: ¥1,950,000 (Direct Cost: ¥1,500,000、Indirect Cost: ¥450,000)
|
|---|
| Keywords | 脳 / 自発活動 / 海馬 / 機械学習 / 人工知能 / 確率 / シナプス可塑性 / 確率的情報処理 / 場所細胞 / 記憶痕跡細胞 / 学習 / 記憶 / 局所回路 / シナプス結合 / ゆらぎ |
|---|
| Outline of Final Research Achievements |
Analyzing simultaneous recordings of a large population of neurons in CA1 of awake animals before, during, and after acquisitions of episodic memories, we revealed various features of neural activities that presumably characterize engram neurons. We also succeeded in developing a biologically plausible learning algorithm of neural networks that utilizing stochastic behaviors of neurons and synapses in the cortical circuit. We found that the learning algorithm consistently accounts for various experimental findings of the brain, solves many known limitations of existing learning rules of neural networks, and provides the most efficient neural coding recently discovered in rodent cortical networks. Our results suggest that synapses and neurons in the cortex cooperatively implement the most efficient learning or stochastic computation.
|
| Academic Significance and Societal Importance of the Research Achievements |
脳の神経細胞及びシナプス結合が示す持続的な確率的挙動に着目することで、ニューラルネットワークに対する脳型の新たな学習アルゴリズムの開発に成功した。このアルゴリズムは既存の学習アルゴリズムの様々な問題点を解決できることが示されており、さらに生物学的妥当性も極めて高いと考えられるため、脳型の人工知能チップや、ニューロモルフィックデバイスの開発などに有用であると期待されるほか、脳の基礎的な動作原理の解明として生命科学にも大きな波及効果を持つと期待される。
|
