Presentation is loading. Please wait.

Presentation is loading. Please wait.

Michale Fee McGovern Institute for Brain Research Department of Brain and Cognitive Sciences MIT Jerusalem in Motion Workshop Jerusalem, Israel December.

Similar presentations


Presentation on theme: "Michale Fee McGovern Institute for Brain Research Department of Brain and Cognitive Sciences MIT Jerusalem in Motion Workshop Jerusalem, Israel December."— Presentation transcript:

1 Michale Fee McGovern Institute for Brain Research Department of Brain and Cognitive Sciences MIT Jerusalem in Motion Workshop Jerusalem, Israel December 18, 2003 Vocal control in the songbird: Neural mechanisms of sequence generation

2 A B C D E F G H I J A-B-C-D-E-F-G-H-I-J

3 123 Abeles, Hertz, ‘80s and ‘90s Synchronous Firing Chain Neural Circuits for Sequence Generation Metastable Attractors Sompolinsky, Kleinfeld, Platt, 1980s fast slow

4 Neural Circuits for Sequence Generation Train a specified sequence of neural states, Sequence of states must be nearly orthogonal A-B-C-A-D is not allowed Interference between sequence and dynamics Timescale is set by synaptic/biophysical time constants W ij =  S i t  S j t +1 SitSit t

5 Overview Songbird as a model system Technological challenges Mechanisms of sequence generation in the songbird

6 Zebra Finches

7 0 kHz 10 kHz Zebra Finch Song Structure 1s Frequency Motif Syllable

8 Songbird Vocalizations are Highly Stereotyped

9 Songbirds Can Generate Output Over a Wide Range of Timescales

10 Biological systems can: Learn and reliably generate low-dimensional sequential behavior –not a specified sequence of neural states Generate an arbitrary sequence – not constrained by orthogonality between output states Operate over a wide range of timescales

11 Circuits for Vocal Production and Learning Motor Circuit Learning Circuit (7) 1000 7000 20,000

12 Technical Difficulties Songbirds will only sing while unconstrained Zebra finch weighs only 12-15 grams Singing is suppressed by handling

13 3 independently controlled electrodes Motorized for remote control 1.5 gram total weight Motorized Miniature Microdrive Fee and Leonardo, 2000

14

15 Premotor Activity During Singing Bout Motif

16

17 Instantaneous Firing Rate 0.00.40.60.80.2 1 6 12 Neuron # Time [s] Firing Rate [1 kHz/Div]

18 How Are the Burst Sequences in RA Generated? Internal dynamics within RA? - OR - Imposed from HVC?

19 Models of Pattern Generation in HVC and RA Feed-forward Intrinsic HVC RA HVC RA ~10ms

20 Singing Related Firing Patterns in Nucleus HVC Yu and Margoliash, 1996

21 Antidromic Identification of HVC Neurons

22 Antidromic identification of HVC neurons

23 What do RA-Projecting HVC neurons do during singing? Hahnloser, Kozhevnikov, and Fee, Nature (2002)

24 Hahnloser, Kozhevnikov and Fee, Nature (2002)

25 Simple Sequence Generation Circuit Sparse representation of time Fixed synaptic weights Plastic synaptic weights

26 Downstream effect of RA activity

27 Simple Sequence Generation Circuit Sparse representation of time Fixed synaptic weights Plastic synaptic weights

28 Model of Vocal Learning with Sebastian Seung and Ila Fiete

29 A Sparse Representation in HVC Speeds Learning with Sebastian Seung and Ila Fiete

30 Simple Sequence Generation Circuit: Emergent RA activity

31 Emergent Activity in RA Neurons with Sebastian Seung and Ila Fiete

32 Emergent Activity in RA Neurons

33 Each model RA neuron has a unique pattern of bursts A different ensemble of active RA neurons at each time in the sequence The ensemble of active RA neurons evolves to an uncorrelated ensemble every ~10 ms, even during constant output

34 RA ensembles are uniquely related to a temporal position in the output – not to motor output How is this possible?

35 High Degree of Convergence From RA to Motor Output ~7000 RA projection neurons ~1000 motor neurons 7 muscles

36 Many Different Ensembles of Active RA Neurons Can Produce the Same Motor Output Model RA outputs form a highly degenerate code for motor signals

37

38 Instantaneous Firing Rate 0.00.40.60.80.2 1 6 12 Neuron # Time [s] Firing Rate [1 kHz/Div]

39 Time t 2 0 200 400 600 125 Neuron # Time t 1 0200400600 1 25 Neuron # Time t 1 Time t 2 0200400600 0 200 400 600

40

41 How are the Timescales of Neural and Motor Activity Related?

42 Neural and Song Correlation Matrices

43 Neural and Song Correlation Width

44 Circuits for Vocal Production and Learning Motor Circuit Learning Circuit

45 Each RA neuron has a unique pattern of bursts A different ensemble of active RA neurons at each time in the song motif The ensemble of active RA neurons evolves to an uncorrelated ensemble every ~10 ms, even during parts of the song with constant acoustic output

46 Our proposed network can: Learn and reliably generate low-dimensional sequential behavior –not a specified sequence of neural states Generate an arbitrary sequence – not constrained by orthogonality between output states Operate over a wide range of timescales

47 Design Principles and Implications Separate the temporal dynamics and the mapping to motor output –Changes in learned output do not affect temporal structure Sparse coding of temporal order in HVC –Fast learning? –No single neuron tuning in RA?

48 Future Directions When during development does the sparse representation of time in HVC arise? Where do sparse sequences in HVC originate? Intrinsic dynamics within HVC, or driven from NIf?

49 Circuits for Vocal Production and Learning Motor Circuit Learning Circuit

50 123 Where and how is ‘time’ generated? fast slow

51 Collaborators Richard Hahnloser –Bell Laboratories Alexay Kozhevnikov –Bell Laboratories Anthony Leonardo –Bell Laboratories Ila Fiete, Sebastian Seung –Brain and cognitive sciences department – MIT

52 Simple Sequence Generation Circuit Hidden Layer Output Sparse representation of time Fixed synaptic weights Plastic synaptic weights

53

54 Simple Models of Neural Circuits stable states - fast, symmetric connections dynamic states - slow or asymmetric connections


Download ppt "Michale Fee McGovern Institute for Brain Research Department of Brain and Cognitive Sciences MIT Jerusalem in Motion Workshop Jerusalem, Israel December."

Similar presentations


Ads by Google