This tool allows the training and testing of a simple MLP (Multilayer Perceptron) network. The specific network is trained, using the MNIST dataset, to be able to classify hand-written digits.
$ make
The dataset folder is required to run the program. This folder contains all the MNIST images and has to be placed into the root folder of the project:
.
├── dataset
│ ├── training
│ │ ├── 0
│ │ ├── 1
│ │ ├── 2
│ │ ├── 3
│ │ ├── 4
│ │ ├── 5
│ │ ├── 6
│ │ ├── 7
│ │ ├── 8
│ │ └── 9
│ └── testing
│ ├── 0
│ ├── 1
│ ├── 2
│ ├── 3
│ ├── 4
│ ├── 5
│ ├── 6
│ ├── 7
│ ├── 8
│ └── 9
├── main.c
├── images_handler.h
├── images_handler.c
├── neural_network_handler.h
├── neural_network_handler.c
├── stb_image.h
└── Makefile
Run the program with:
$ ./MNIST_classification_in_C -l 1 -n 64 -e 10 -lr 0.001
where:
- The
-largument sets the number of hidden layers. - The
-nargument sets the number of neurons of each hidden layer. - The
-eargument sets the number of training epochs. - The
-lrargument sets the specific learning rate used during training.
Reading ./dataset/training/0/ files.
Reading ./dataset/training/1/ files.
Reading ./dataset/training/2/ files.
Reading ./dataset/training/3/ files.
Reading ./dataset/training/4/ files.
Reading ./dataset/training/5/ files.
Reading ./dataset/training/6/ files.
Reading ./dataset/training/7/ files.
Reading ./dataset/training/8/ files.
Reading ./dataset/training/9/ files.
Reading ./dataset/testing/0/ files.
Reading ./dataset/testing/1/ files.
Reading ./dataset/testing/2/ files.
Reading ./dataset/testing/3/ files.
Reading ./dataset/testing/4/ files.
Reading ./dataset/testing/5/ files.
Reading ./dataset/testing/6/ files.
Reading ./dataset/testing/7/ files.
Reading ./dataset/testing/8/ files.
Reading ./dataset/testing/9/ files.
Shuffling samples.
Training.
Epoch: 1/10, accuracy: 0.759783, loss: 1.746183
Epoch: 2/10, accuracy: 0.861150, loss: 0.689133
Epoch: 3/10, accuracy: 0.883467, loss: 0.566282
Epoch: 4/10, accuracy: 0.896183, loss: 0.498846
Epoch: 5/10, accuracy: 0.905183, loss: 0.453839
Epoch: 6/10, accuracy: 0.912250, loss: 0.420292
Epoch: 7/10, accuracy: 0.917500, loss: 0.393635
Epoch: 8/10, accuracy: 0.922050, loss: 0.371836
Epoch: 9/10, accuracy: 0.925733, loss: 0.353492
Epoch: 10/10, accuracy: 0.929567, loss: 0.337611
Testing.
Accuracy: 0.926500, loss: 0.354491
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent