learning rule
Perceptron — update only on mistakes
Delta rule — update on every example
Logistic regression — cross-entropy
learning rate η —
0.05
class overlap — none … large (
0.00
)
flipped labels —
none
next example
run
one pass
reset w
new data
pause on mistakes
examples seen
0
mistakes so far
0
updates this pass
–
total
0
pass
0
training accuracy
–
length of w
0