Page 20 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

4. Robert Lucky’s Adaptive Equalization, From the 1960s 7

(A)

(B)
k e

Positive stable
+ + equilibrium point
y k
– – ( SUM )
Negative stable
Unstable
equilibrium point
equilibrium point

FIGURE 1.4
Bootstrap learning. (A) The quantized output, the sum, and the error as a function of
(SUM). (B) The error versus (SUM).

4. ROBERT LUCKY’S ADAPTIVE EQUALIZATION, FROM THE
1960s
In the early 1960s, as Widrow’s group at Stanford was developing bootstrap
learning, at the same time, independently, a project at Bell laboratories led by Robert
W. Lucky was developing an adaptive equalizer for digital data transmission over
telephone lines [10,11]. His adaptive algorithm incorporated what he called “deci-
sion directed learning,” which has similarities to bootstrap learning.