untitled

Size: px

Start display at page:

Download "untitled"

もりよりさくいし
7 years ago
Views:

1 Discovery Channel Biped Schaal S, Sternad D, Osu R, Kawato M: Rhythmic arm movement is not discrete. Nature Neuroscience, 7, (004). Nakanishi J, Morimoto J, Endo G, Cheng G, Schaal S, Kawato M: Learning from demonstration and adaptation of biped locomotion, J. Robotics and Autonomous Systems, 47, (004).

Nakanishi J, Morimoto J, Endo G, Cheng G, Schaal S, Kawato M: Learning from

2 CB-i 155cm, 85kg PC PC JST()Duke() CB-i Duke()JST()

4 Parallel Fiber and Climbing Fiber Inputs to Purkinje Cells induce Simple Spikes and Complex Spikes A Parallel fiber (PF) B Simple spike (SS) Internal model output Input signal Purkinje cell Climbing fiber (CF) Complex Complex spike spike (CS)(CS) Error signal Error signal Output signal Error signal 1 = (M L 1 + M L 1 S cos + I 1 + I ) 1 + (M L 1 S cos + I ) M L 1 S ( 1 + ) sin + B 1 1 = (M L 1 S cos + I ) 1 + I + M L 1 S 1 sin + B Jun Nakanishi, and Stefan SchaalFeedback error learning and nonlinear adaptive control. Neural Networks, 17, (004)

signal Error signal 1 = (M L 1 + M L 1 S cos + I 1 + I ) 1 + (M L 1 S cos + I ) M L 1 S ( 1 + ) sin + B 1 1 = (M L 1 S cos + I ) 1 + I + M L 1 S

5 d dt = ( ff ) T fb desired E = 1 ( desired ff ) T ( desired ff ) d dt = ( ff ) T ( desired ff ) fb ( desired ff ) fb desired y n x i i y = i x i n d i dt = x i (C C spont ) C C spont d i dt = ( ff i ) fb = ((y) i ) fb = x i (C C spont ) Simple and Complex Spikes of Purkinje Cells in Monkeys during Ocular Following Responses Kawano, Shidara, Gomi, Kobayashi, Takemura Ocular Following Responses: Reflex eye movement induced by movement of large visual field f()= t M ( t + )+ B ( t + )+ K( t + )+ f bias

Complex Spikes of Purkinje Cells in Monkeys during Ocular Following Responses Kawano, Shidara, Gomi, Kobayashi, Takemura

6 イメージを表示で A B C D F G H Spikes/s OFRVPFL Winkelman & Frens (005) Lange E q p p 回転変換 q y y x x cos sin p = y sin cos q 10 x 積分変換 x = p y q

7 r rio e p u S ) m (m Z r rio fe In Left Imamizu H, Miyauchi S, Tamada T, Sasaki Y, Takino R, Puetz B, Yoshioka T, Kawato M: Human cerebellar activity reflecting an acquired internal model of a new tool. Nature (000) X (mm) Right Posterior Anterior Y (mm) Organization of 16 Daily Tools in the Right Superior Cerebellum MRI Mirror Medial Higuchi et al. Cortex (007) Ant. lateral Language and Tool Internal Models in Broca s Area: Segregation and Overlap 8 healthy subjects ( 14 males and 14 females) Conditions Tool-use execution Random effect analysis 16 tools and utensils: Chopsticks, saw, scissors, pencil, hammer, screw-driver, fork, spoon, tooth brush, brush, comb, cutting pliers, monkey wrench, wrench, knife and clip Higuchi et al. Cortex (007) Averaged across 8 subjects with fixed effect model p<0.05 corrected. The peaks are projected in a horizontal plane Conditions Tool-use execution Hold the tool and look at the object control Tool-use motor imagery Pos. Hold the tool and look at the object control Tool-use motor imagery Story listening Reversed story listening (control) Higuchi, Chaminade, Imamizu, Kawato (in preparation) Typical subject A mc A x A mc B x B B,mc,mc Q * mc B = mc A +(1 )mc A 0

lateral Language and Tool Internal Models in Broca s Area: Segregation and Overlap 8 healthy subjects ( 14 males and 14 females) Conditions Tool-use execution Random effect analysis 16 tools and

8 () BMI 0 40 V 5 s BMI 1.. SVM, SLR, SR 3. Bayes ()

9 P(W n ) = a n exp 1 a W n n P(W n ) = a n exp 1 a (W n n μ) ブレインネットワークインタフェースの利点逆問題解法と特徴量絞り込みの組み合わせ sensor signals 400 brain activity 1,00010,000 X Y F (lead field matrix) X Y A (VB filter) g physics or mental sate Z= g (Y) Z = g (AX) = h (X) Equivalent nonsense But sparsenes makes big difference Z Ting Brute & SR Neuron Regression Yamashita SLR fmri Classification Ganesh LASSO fmri Regression Toda SR MEG & fmri (VB) Regression Nambu Brute & SR NIRS Regression Okabe SLR EEG & fmri (VB) Classification Z vbmeg MEG topography Averaging MEG signal Time current density Time Toda A, Imamizu H, Sato M, Wada Y, Kawato M: Reconstruction of temporal movement from single-trial non-invasive brain activity: A hierarchical Bayesian method. Proceedings of 14th International Conference on Neural Information Processing(ICONIP 007). WED-4, p.131 (007)

jp/ physics or mental sate Z= g (Y) Z = g (AX) = h (X) Equivalent nonsense But sparsenes makes big difference Z Ting Brute & SR Neuron Regression Yamashita SLR fmri Classification Ganesh LASSO fmri

10 X SPM Reconstructions Dynamics Motor control Sensory feedback Hierarchical Reinforcement Learning Partial and incomplete observation, embedding necessary, order parameter, high-level description, Kernel CCA Y CCA in neuroscience context Sawahata, Fujiwara, Kamitani, Suetani, entirely independent Towards New Paradigm Conventional approach Correlation Information not directly manipulated Reductionism Hypothesis driven Hard to bridge different hierarchies New direction Prediction of whole dynamics of brain, mind and world Information manipulated Synthesis Data driven Bridge hierarchies Microcircuitry of Cerebello-IO Network De Zeeuw CI, Simpson JI, Hoogenraad CC, Galjart N, Koekkoek SKE, Ruigrok TJH: Microcircuitry and function of the inferior olive. Trends in Neurosciences., 1, (1998). Controversies of Inferior Olive Functions: Rhythmicity and Synchrony versus Learning Signals 1. Most intensive gap junctions (electrical coupling) between IO neurons. Strong rhythmicity and synchrony under anesthetized rodents with blockades of synaptic inputs to IO (Llinas) 3. No rhythmicity and little synchrony for awake monkeys (Thach) 4. Leaning theory should explain gap junctions and low firing rates Chaotic Resonance in Inferior Olive Biophysical model of IO neuron Schweighofer N, Doya K, Kawato M: Electrophysiological propersties of infereor olive neurons: a compartmental model. Journal of Neurophysiology 8, (1999). Schweighofer N, Doya K, H. Fukai, Chiron JV, Furukawa T, Kawato. M: Chaos may enhance information transmission in the inferior olive. Proc Natl Acad Sci USA., 101, (004). Same initial condition Intermediate coupling Strong coupling

Reductionism Hypothesis driven Hard to bridge different hierarchies New direction Prediction of whole dynamics of brain, mind and world Information manipulated Synthesis Data driven Bridge

11 Spiking of 3x3 Cells without Inputs Largest Lyapunov Exponent Network Mutual Informatio Conclusions of IO Chaotic Resonance 1. Chaotic resonance can enhance information transmission with low firing frequency.. Chaos is more efficient than synaptic noise in information transmission. 3. IO network model can explain decrease of rhythmicity and synchrony with decrease in coupling under physiological synaptic inputs. 4. Partial synchrony of a small subset of IO cells was reproduced Picrotoxin condition (1) GABAA Gap junction Gap junction Picrotoxin condition coupling noise GABAA Picrotoxin condition () Picrotoxin condition (gc=0.6) E. J. Lang et al. (1996) 00 Auto-correlogram 00 Cross-correlogram ms

IO network model can explain decrease of rhythmicity and synchrony with decrease in coupling under physiological synaptic inputs. 4.

12 noise Control condition (gc=0.05) Coupling conductance=0.05 (mv) 0 Auto-correlogram 0 Cross-correlogram ms (ms)

21 David Marr Marr Marr Marr 3 1. 1

21 David Marr Marr Marr Marr 3 1. 1 2 2. 2.1. 3.1.1. 3 (1) (2) () (4) (5) 3.1.2. 3.1.4. 1970 1984 Doya K. What are the computations of the cerebellum, the basal ganglia, and the cerebral cortex. Neural