3 Layer NN with smaller layer size

[rhemon]

12 - 12 - 4
0.02 learning rate
3000 iterations

Accuracy with training set, 64.328125%
Accuracy with dev set, 68.75%
Accuracy in recognising mango: 100.0
Accuracy in recognising apple: 50.0
Accuracy in recognising orange: 50.0
Accuracy in recognising pear: 50.0

CircleCI Log

[rhemon]

12 - 12 - 4
0.05 learning rate
3000 iterations

Accuracy with training set, 64.453125%
Accuracy with dev set, 68.0%
Accuracy in recognising mango: 52.3333333333
Accuracy in recognising apple: 97.7777777778
Accuracy in recognising orange: 50.0
Accuracy in recognising pear: 50.0

CircleCI Log

[rhemon]

12 - 12 -4
0.1 learning rate
3000 iterations

Accuracy with training set, 64.703125%
Accuracy with dev set, 69.25%
Accuracy in recognising mango: 50.0
Accuracy in recognising apple: 51.8888888889
Accuracy in recognising orange: 99.6666666667
Accuracy in recognising pear: 50.0

CircleCI Log

[rhemon]

12 - 12 - 4
0.9 learning rate
3000 iterations

Accuracy with training set, 64.765625%
Accuracy with dev set, 69.25%
Accuracy in recognising mango: 50.0
Accuracy in recognising apple: 52.1111111111
Accuracy in recognising orange: 99.6666666667
Accuracy in recognising pear: 50.0

CircleCI Log

[rhemon]

12 - 12 - 4
1.5 learning rate
3000 iterations

Accuracy with training set, 64.828125%
Accuracy with dev set, 69.25%
Accuracy in recognising mango: 50.0
Accuracy in recognising apple: 52.3333333333
Accuracy in recognising orange: 99.6666666667
Accuracy in recognising pear: 50.0

CircleCI Log