CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

14
CS 621 Artificial Intelligence Lecture 33 - 07/11/05 Guest Lecture by Prof. Rohit Manchanda Biological Neurons - I

description

The Perceptron Model A perceptron is a computing element with input lines having associated weights and the cell having a threshold value. The perceptron model is motivated by the biological neuron. Output = y Threshold = θ w1 wn Wn-1 x1 Xn-1

Transcript of CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Page 1: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

CS 621 Artificial Intelligence

Lecture 33 - 07/11/05

Guest Lecture byProf. Rohit Manchanda

Biological Neurons - I

Page 2: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

The Perceptron Model A perceptron is a computing element with input lines having associated weights and the cell having a threshold value. The perceptron model is motivated by the biological neuron.

Output = y

wnWn-1

w1

Xn-1

x1

Threshold = θ

Page 3: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Step function / Threshold functiony = 1 for Σ > θ, y >1

=0 otherwise

Σwixi > θ

θ

1y

Σwixi

Features of Perceptron

• Input output behavior is discontinuous and the derivative does not exist at Σwixi = θ • Σwixi - θ is the net input denoted as net

• Referred to as a linear threshold element - linearity because of x appearing with power 1

• y= f(net): Relation between y and net is non-linear

Page 4: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Perceptron vs. the point neuron

• Incoming signals from synapses are summed up at the soma

• , the biological “inner product”• On crossing a threshold, the cell “fires”

generating an action potential in the axon hillock region

The McCulloch and Pitt’s neuron

Synaptic inputs: Artist’s conception

Page 5: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

The biological neuron

Pyramidal neuron, from the amygdala (Rupshi et al. 2005)

A CA1 pyramidal neuron (Mel et al. 2004)

Page 6: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

What makes neuronal computation special ?

• Electrical activity in neurons• Traveling action potentials• Leaky, conducting membrane• Highly branched structures• A variety of electrical connections between neurons• A varied range of possibilities in neural coding of

information

Page 7: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Dendritic computation

• Time delay• Attenuation, duration increase• Spatial and temporal

summation• Non linear intra branch

summation• Multiple layer configuration• Retrograde propagation of

signals into the dendritic arbor– Backprogating action

potentials

Spatial and temporal summation of proximal and distal inputs (Nettleton et al. 2000)

Page 8: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Dendritic computationComparison of within-branch andbetween-branch summation.

(a-f)Black: individually Activated Re: simultaneously activatedBlue: arithmetic sum of individual responses.

(g) Summary plot of predicted versus actual combined responses. Coloured circles: within-branch summationDashed line: linear summation. Green diamonds: between-branch summation

(h) Modeling data: summation of EPSPsRed circles: within-branch summation Open green circles: between-branch summation

(Mel et al. 2003, 2004)

Page 9: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Possible computational consequences of non-linear summation in dendritic sub-trees (Mel et al. 2003)

Dendritic computation: Single neuron as a neural

network

Another possibility

Page 10: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Single neuron as a neural network

3 Layered Neural Network2 Layered Neural Network

Page 11: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

A multi neuron pathway in the hippocampus

Page 12: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Summary

Classical Picture Emerging Picture

Page 13: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Recent neurobiological discoveries

Changes as a result of stress (CIS)– Dendritic atrophy

Consequences– Loss of computational subunits– Changes in connectivity of the

network

Atrophy and de-branching as seen in the CA3 pyramidal cells from (Ajai et al. 2002)

Page 14: CS 621 Artificial Intelligence Lecture /11/05 Guest Lecture by Prof

Increase in spine count (Amygdaloid neurons) (Rupshi et al. 2005)

Changes as a result of stress (CIS)

•Arbor growth•Increase in spine count

Consequences

•Increase in computational subunits•Synaptic site increase

Recent neurobiological discoveries