ABCDEFGHIJKLMNOPQRSTUVWXYZAAAB
1
2
3
Iris Classification with Google Sheets
4
5
Alexander Gurung, David Zhao, Naitian Zhou
6
7
This is a neural network classifier that classifies the classic Iris species dataset with a 95% accuracy.
And it was written, trained, and validated all a Google spreadsheet document.
8
9
The iris sheet is the training data.
10
The mat_neural_net sheet is our original (failed) implementation
11
The mat_neural_net_v2 sheet is our new (working) implementation
12
E2-10 are the training epochs. Epoch 1 is in the mat_neural_net_v2 sheet.
Note: I hardcoded the weights at the beginning of each epoch using the weights at the end of the previous epoch so Google sheets has less calculation to do on each page load. However, since it does still recalculate each individual epoch, these weights may no longer line up.
13
Accuracy contains the accuracy measurement (using the final weights at the end of the 10th epoch)
14
error contains the graph of error across iterations
15
hyperparameters contains the hyperparameters that we used, along with some other random stuff.
16
17
Architecture
18
19
We ended up using a fully connected neural network with no hidden layers.
20
The output layer had used a softmax activation function.
21
The loss was calculated with cross entropy.
22
23
Our old architecture used a hidden layer with 3 hidden nodes. Both the hidden and output nodes had a sigmoid activation
24
function. Sigmoid activations suck. Don't ever use them.
25
26
27
About the Team
28
29
Naitian Zhou is a senior at Thomas Jefferson High School for Science and Technology.
30
31
Alex Gurung is a senior at Thomas Jefferson High School for Science and Technology.
32
33
David Zhao is a sophomore at the University of Virginia.
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100