mattdibi / myANN

My Artificial Neural Network

myANN

My Artificial Neural Network.

I used this project to learn OpenCV basics and to have a better understanding of how an Artificial Neural Network works. The network characteristics:

• Forward propagation
• Optimization: backpropagation algorithm w/ gradient descent

Requirements

• cmake
• OpenCV

Results

Specs:

• CPU: Intel Core i7 4790K (4 cores/8 threads @ 4.4GHz)
• RAM: 16GB 1600MHz

Test 1

• Training set size = 25000 samples
• Learning rate = 0.75

Correction rate: 90.68%
Cost function output: 0.643168

Training set size: 25000
myANN Settings:
- Activation function: SIGMOIDAL
- Max Iterations: 350
- Learning Rate: 0.75
- Number of layers: 3
-  Layer 0 layer dimension: 784
-  Layer 1 layer dimension: 200
-  Layer 2 layer dimension: 10
- Number of matrices: 2
-  Layer 0 matrix dimension: 785x200
-  Layer 1 matrix dimension: 201x10

real	9m42.199s
user	70m53.180s
sys	0m3.432s

Test 2

• Training set size = 18000 samples
• Learning rate = 0.1

Correction rate: 80.65%
Cost function output: 1.22711

Training set size: 18000
myANN Settings:
- Activation function: SIGMOIDAL
- Max Iterations: 350
- Learning Rate: 0.1
- Number of layers: 3
-  Layer 0 layer dimension: 784
-  Layer 1 layer dimension: 200
-  Layer 2 layer dimension: 10
- Number of matrices: 2
-  Layer 0 matrix dimension: 785x200
-  Layer 1 matrix dimension: 201x10

real	7m22.989s
user	51m49.860s
sys	0m3.460s