Refactor to use neural network instead of linear classifier

[jpsember]

http://cs231n.github.io/neural-networks-case-study/

[jpsember]

Ok, I've written the code and got it running... maybe I need to increase the number of training samples.

Rep:  1 Cost:   11.316 Speed:    0.300
Rep:  2 Cost:   12.479 Speed:    0.300
Rep:  3 Cost:    7.709 Speed:    0.300
Rep:  4 Cost:    8.812 Speed:    0.300
Rep:  5 Cost:    7.553 Speed:    0.300
Rep:  6 Cost:    6.289 Speed:    0.300
Rep:  7 Cost:    6.596 Speed:    0.300
Rep:  8 Cost:    5.780 Speed:    0.300
Rep:  9 Cost:    5.032 Speed:    0.300
Rep: 10 Cost:    3.357 Speed:    0.300
Rep: 11 Cost:    3.067 Speed:    0.300
Rep: 12 Cost:    2.896 Speed:    0.300
Rep: 13 Cost:    3.817 Speed:    0.300
Rep: 14 Cost:    2.830 Speed:    0.300
Rep: 15 Cost:    3.269 Speed:    0.300
Rep: 16 Cost:    2.130 Speed:    0.300
Rep: 17 Cost:    2.024 Speed:    0.300
Rep: 18 Cost:    1.813 Speed:    0.300
Rep: 19 Cost:    2.113 Speed:    0.300
Rep: 20 Cost:    1.718 Speed:    0.300
Rep: 21 Cost:    1.989 Speed:    0.300
Rep: 22 Cost:    1.800 Speed:    0.300
Rep: 23 Cost:    1.689 Speed:    0.300
Rep: 24 Cost:    1.307 Speed:    0.300
Rep: 25 Cost:    1.114 Speed:    0.300
Rep: 26 Cost:    1.080 Speed:    0.300
Rep: 27 Cost:    1.117 Speed:    0.300
Rep: 28 Cost:    1.135 Speed:    0.300
Rep: 29 Cost:    1.099 Speed:    0.300
Rep: 30 Cost:    0.945 Speed:    0.300
Rep: 31 Cost:    0.906 Speed:    0.300
Rep: 32 Cost:    0.924 Speed:    0.300
Rep: 33 Cost:    0.940 Speed:    0.300
Rep: 34 Cost:    0.850 Speed:    0.300
Rep: 35 Cost:    0.892 Speed:    0.300
Rep: 36 Cost:    0.810 Speed:    0.300
Rep: 37 Cost:    0.862 Speed:    0.300
Rep: 38 Cost:    0.862 Speed:    0.300
Rep: 39 Cost:    0.889 Speed:    0.300
Rep: 40 Cost:    0.838 Speed:    0.300
Rep: 41 Cost:    0.815 Speed:    0.300
Rep: 42 Cost:    0.704 Speed:    0.300
Rep: 43 Cost:    0.744 Speed:    0.300
Rep: 44 Cost:    0.722 Speed:    0.300
Rep: 45 Cost:    0.724 Speed:    0.300
Rep: 46 Cost:    0.689 Speed:    0.300
Rep: 47 Cost:    0.670 Speed:    0.300
Rep: 48 Cost:    0.636 Speed:    0.300
Rep: 49 Cost:    0.687 Speed:    0.300
Rep: 50 Cost:    0.672 Speed:    0.300
Accuracy:     0.833

[jpsember]

Rep:  1 Cost:   18.987 Speed:    0.700
Rep:  2 Cost:   42.524 Speed:    0.700
Rep:  3 Cost:    3.918 Speed:    0.700
Rep:  4 Cost:    6.117 Speed:    0.700
Rep:  5 Cost:   11.945 Speed:    0.700
Rep:  6 Cost:    4.031 Speed:    0.700
Rep:  7 Cost:    1.821 Speed:    0.700
Rep:  8 Cost:    1.428 Speed:    0.700
Rep:  9 Cost:    1.188 Speed:    0.700
Rep: 10 Cost:    1.034 Speed:    0.700
Rep: 11 Cost:    0.917 Speed:    0.700
Rep: 12 Cost:    0.836 Speed:    0.700
Rep: 13 Cost:    0.776 Speed:    0.700
Rep: 14 Cost:    0.732 Speed:    0.700
Rep: 15 Cost:    0.697 Speed:    0.700
Rep: 16 Cost:    0.683 Speed:    0.700
Rep: 17 Cost:    0.707 Speed:    0.700
Rep: 18 Cost:    0.765 Speed:    0.700
Rep: 19 Cost:    0.963 Speed:    0.700
Rep: 20 Cost:    0.682 Speed:    0.700
Rep: 21 Cost:    0.627 Speed:    0.250
Rep: 22 Cost:    0.610 Speed:    0.250
Rep: 23 Cost:    0.601 Speed:    0.250
Rep: 24 Cost:    0.594 Speed:    0.250
Rep: 25 Cost:    0.588 Speed:    0.250
Rep: 26 Cost:    0.582 Speed:    0.250
Rep: 27 Cost:    0.577 Speed:    0.250
Rep: 28 Cost:    0.573 Speed:    0.250
Rep: 29 Cost:    0.568 Speed:    0.250
Rep: 30 Cost:    0.564 Speed:    0.250
Rep: 31 Cost:    0.561 Speed:    0.100
Rep: 32 Cost:    0.559 Speed:    0.100
Rep: 33 Cost:    0.558 Speed:    0.100
Rep: 34 Cost:    0.556 Speed:    0.100
Rep: 35 Cost:    0.555 Speed:    0.100
Rep: 36 Cost:    0.553 Speed:    0.100
Rep: 37 Cost:    0.552 Speed:    0.100
Rep: 38 Cost:    0.551 Speed:    0.100
Rep: 39 Cost:    0.549 Speed:    0.100
Rep: 40 Cost:    0.548 Speed:    0.100
Rep: 41 Cost:    0.547 Speed:    0.100
Rep: 42 Cost:    0.545 Speed:    0.100
Rep: 43 Cost:    0.544 Speed:    0.100
Rep: 44 Cost:    0.543 Speed:    0.100
Rep: 45 Cost:    0.541 Speed:    0.100
Rep: 46 Cost:    0.540 Speed:    0.100
Rep: 47 Cost:    0.539 Speed:    0.100
Rep: 48 Cost:    0.537 Speed:    0.100
Rep: 49 Cost:    0.536 Speed:    0.100
Rep: 50 Cost:    0.535 Speed:    0.100
Rep: 51 Cost:    0.533 Speed:    0.100
Rep: 52 Cost:    0.532 Speed:    0.100
Rep: 53 Cost:    0.531 Speed:    0.100
Rep: 54 Cost:    0.530 Speed:    0.100
Rep: 55 Cost:    0.528 Speed:    0.100
Rep: 56 Cost:    0.527 Speed:    0.100
Rep: 57 Cost:    0.526 Speed:    0.100
Rep: 58 Cost:    0.525 Speed:    0.100
Rep: 59 Cost:    0.524 Speed:    0.100
Rep: 60 Cost:    0.523 Speed:    0.100
Rep: 61 Cost:    0.521 Speed:    0.100
Rep: 62 Cost:    0.520 Speed:    0.100
Rep: 63 Cost:    0.519 Speed:    0.100
Rep: 64 Cost:    0.518 Speed:    0.100
Rep: 65 Cost:    0.517 Speed:    0.100
Rep: 66 Cost:    0.516 Speed:    0.100
Rep: 67 Cost:    0.515 Speed:    0.100
Rep: 68 Cost:    0.514 Speed:    0.100
Rep: 69 Cost:    0.513 Speed:    0.100
Rep: 70 Cost:    0.512 Speed:    0.100
Rep: 71 Cost:    0.511 Speed:    0.100
Rep: 72 Cost:    0.510 Speed:    0.100
Rep: 73 Cost:    0.509 Speed:    0.100
Rep: 74 Cost:    0.508 Speed:    0.100
Rep: 75 Cost:    0.507 Speed:    0.100
Accuracy:     0.963
py> 

[jpsember]