▶ Start / restart
iteration
0
/100
· avg error
50.0%
learning rate 2.0 · 100 iterations
smile% = σ(score) = 1 ⁄ (1 + e
−score
)
·
score = w
1
·p
1
+ w
2
·p
2
+ … + w
12
·p
12
the 12 multipliers
0
0
0
0
0
0
0
0
0
0
0
0
training faces
target
100%
score
0
σ →
50%
target
0%
score
0
σ →
50%
target
0%
score
0
σ →
50%
target
100%
score
0
σ →
50%
loss — average error
0%
25%
50%
draw a face · test it
σ
50%
(0)
→
smile
✕ clear
predicting with: trained · 100 iterations