basics of computational neuroscience - uni-goettingen.de · basics of computational neuroscience. 2...
TRANSCRIPT
1
Basics of ComputationalNeuroscience
2
Lecture: Computational Neuroscience, Contents
1) IntroductionThe Basics – A reminder:
1) Brain, Maps, Areas, Networks, Neurons, and SynapsesThe tough stuff:
2,3) Membrane Models3,4) Spiking Neuron Models5) Calculating with Neurons I: adding, subtracting, multiplying, dividing5,6) Calculating with Neurons II: Integration, differentiation6) Calculating with Neurons III: networks, vector-/matrix- calculus, assoc. memory6,7) Information processing in the cortex I: Neurons as filters7) Information processing in the cortex II:Correlation analysis of neuronal connections7,8) Information processing in the cortex III: Neural Codes and population responses8) Information processing in the cortex IV: Neuronal maps
Something interesting – the broader perspective9) On Intelligence and Cognition – Computational Properties?
Motor Function10,11) Models of Motor Control
Adaptive Mechanisms11,12) Learning and plasticity I: Physiological mechanisms and formal learning rules12,13) Learning and plasticity II: Developmental models of neuronal maps13) Learning and plasticity III: Sequence learning, conditioning
Higher functions14) Memory: Models of the Hippocampus15) Models of Attention, Sleep and Cognitive Processes
3
Literature (all of this is very mathematical!)
General Theoretical Neuroscience:
„Theoretical Neuroscience“, P.Dayan and L. Abbott, MIT Press (there used to be a version of this on theinternet)
„Spiking Neuron Models“, W. Gerstner & W.M. Kistler, Cambridge University Press. (there is a version on the internet)
Neural Coding Issues: „Spikes“ F. Rieke, D. Warland, R. de Ruyter v. Steveninck, W. Bialek, MIT Press
Artificial Neural Networks: „Konnektionismus“, G. Dorffner, B.G. Teubner Verlg. Stuttgart
„Fundamentals of Artificial Neural Networks“, M.H. Hassoun, MIT Press
Hodgkin Huxley Model: See above „Spiking Neuron Models“, W. Gerstner & W.M. Kistler, Cambridge University Press.
Learning and Plasticity: See above „Spiking Neuron Models“, W. Gerstner & W.M. Kistler, Cambridge University Press.
Calculating with Neurons: Has been compiled from many different sources.
Maps: Has been compiled from many different sources.
4
What is computational neuroscience ?
The Interdisciplinary Nature of Computational Neuroscience
5
Neuroscience:
Environment
Stimulus
Behavior
Reaction
Different Approaches towards Brain and Behavior
6
Psychophysics (human behavioral studies):
Environment
Stimulus
Behavior
Reaction
7
Environment
Stimulus
Neurophysiology:
Behavior
Reaction
8
Environment
Stimulus
Theoretical/Computational Neuroscience:
Behavior
Reaction dx
)(xf
U
9
Levels of information processing in the nervous system
Molecules0.1m
Synapses1m
Neurons100m
Local Networks1mm
Areas / „Maps“1cm
Sub-Systems10cm
CNS1m
10
CNS (Central Nervous System):
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
11
Cortex:
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
12
Where are things happening in the brain.
Is the informationrepresented locally ?
The Phrenologists viewat the brain(18th-19th centrury)
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
13
Results from human surgery
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
14
Results from imaging techniques – There are maps in the brain
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
15
Visual System:More than 40 areas !
Parallel processing of „pixels“ andimage parts
Hierarchical Analysis of increasinglycomplex information
Many lateral and feedbackconnections
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
16
Primary visual Cortex:
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
17
Retinotopic Maps in V1: V1 contains a retinotopic map of the visual Field. Adjacent Neurons represent adjacent regions in theretina. That particular small retinal region from which a single neuron receives its input is called the receptivefield of this neuron.
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
V1 receives information from both eyes. Alternatingregions in V1 (Ocular Dominanz Columns) receive(predominantely) Input from either the left or the right eye.
Each location in the cortex represents a different partof the visual scene through the activity of manyneurons. Different neurons encode different aspects of the image. For example, orientation of edges, color, motion speed and direction, etc.
V1 decomposes an image into these components.
18
Orientation selectivity in V1:
Orientation selective neurons in V1 changetheir activity (i.e., their frequency forgenerating action potentials) depending on the orientation of a light bar projected ontothe receptive Field. These Neurons, thus, represent the orientation of lines oder edgesin the image.
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
Their receptive field looks like this:
stimulus
19
Superpositioning of maps in V1: Thus, neurons in V1 are orientationselective. They are, however, also selective for retinal position and ocular dominance as well as forcolor and motion. These are called„features“. The neurons aretherefore akin to „feature-detectors“.
For each of these parameter thereexists a topographic map.
These maps co-exist and aresuperimposed onto each other. In this way at every location in thecortex one finds a neuron whichencodes a certain „feature“. Thisprinciple is called „full coverage“.
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
20
Local Circuits in V1:
Selectivity is generated byspecific connections
stimulus
Orientation selectivecortical simple cell
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
21
Layers in the Cortex:
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
22
Local Circuits in V1:
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
LGN inputs Cell types
Spiny stellatecell Smooth stellate
cell
Circuit
23
Considerations for a Cortex Model• Input
– Structure of the visual pathway• Anatomy of the Cortex
– Cell Types– Connections
• Topography of the Cortex– „X-Y Pixel-Space“ and its distortion– Ocularity-Map– Orientation-Map– Color
• Functional Connectivity of the cortex– Connection Weights– Physiological charateristics of the neurons
At least all these things need to beconsidered when making a „complete“cortex model
24
At the dendrite the incomingsignals arrive (incoming currents)
Molekules
Synapses
Neurons
Local Nets
Areas
Systems
CNS
At the soma currentare finally integrated.
At the axon hillock action potentialare generated if the potential crosses the membrane threshold
The axon transmits (transports) theaction potential to distant sites
At the synapses are the outgoingsignals transmitted onto the dendrites of the targetneurons
Structure of a Neuron:
25
Different Types of Neurons:
Unipolarcell
Bipolarcell
(Invertebrate N.) Retinal bipolar cell
dendrite
axon
soma
Spinal motoneuronHippocampalpyramidal cell
Purkinje cell of thecerebellum
axonsoma
dendrite
Different Typesof Multi-polarCells
26rest
Cell membrane:
K+
Cl-
Ion channels:
Membrane - Circuit diagram: