Lecture 8 - Neural Networks

Question 1

strength of synapse has what effect?

Answer 1

stronger = bigger effect of each incoming pulse on the pulse rate of the downstream cell

Question 2

how do we model neurons?

Answer 2

weighted sum -> non-linearity

Question 3

how to model non-linearity?

Answer 3

use sigmoid function

Question 4

sigmoid function is ___ increasing

Answer 4

monotonic

Question 5

what do we use to get sigmoid function?

Answer 5

hyperbolic tangent equation

Question 6

why do we use nonlinear functions to model neurons?

Answer 6

1. realistic
2. combine into versatile networks (perform any function, not just linear ones)

Question 7

how is the function that the nework computes changed?

Answer 7

function depends on WEIGHT

Question 8

networks of sigmoid cells are _____ approximators. What does this mean?

Answer 8

universal;

these networks can compute any continuous function, given enough cells

Question 9

how to find out how many synapses there are in layers 1, 2, and 3?

e.g. 8-3-5

Answer 9

multiply layer 1 x 2

multiply layer 2 x 3

add results together

e.g. 24 + 15 = 39