/
README.txt
46 lines (37 loc) · 1.87 KB
/
README.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
This is a supervised learning project which conducts 3 types of experiments.
The Dataset is from UC Irvine Machine Learning Repository. This can be found at
https://archive.ics.uci.edu/ml/datasets/Optical+Recognition+of+Handwritten+Digits
with the data being found at
https://archive.ics.uci.edu/ml/machine-learning-databases/optdigits/
Experiment 1: Compares Sum Of Squares error function to Cross entropy error function using different types of hyper
parameters using ReLU as the activation function for the hidden layers and a softmax function for the output layer.
For the best model it outputs confusion matrix.
Experiment 2: Compares tanh vs ReLU as the activation function for the hidden units.
Experiment 3: using a cross-entropy error function and ReLU activation function for hidden units calculate loss and
accuracy using a convolutional neural network.
This project uses Python 3.6.7 with the following libraries
Keras-Applications 1.0.7 1.0.7
matplotlib 3.0.3 3.0.3
numpy 1.16.2 1.16.2
pandas 0.24.2 0.24.2
pip 10.0.1 19.0.3
seaborn 0.9.0 0.9.0
tensorflow 1.13.1 1.13.1
scikit-learn 0.20.3 0.20.3 (this is just used to create validation data. This can be done manually)
tensorflow-gpu 1.13.1 1.13.1 (for convolutional neural network)
If you do not have a GPU you may want to comment out any code related to convolutional networks
The full set of experiments took a total of to run using 431.165 seconds a i7-4790k and a GTX 980.
Instructions for Running
Ensure all libraries are installed as listed above.
Project Structure is as follows:
NeuralNetworkClassifier
Data
test
optdigits.tes
train
examples.csv
labels.csv
optdigits.tra
optdigits-orig.names.txt
main.py
Run main.py