This project implements a minimal neural network in C for classifying handwritten digits from the MNIST dataset. The entire implementation is ~200 lines of code and uses only the standard C library.
- Two-layer neural network (input → hidden → output)
- ReLU activation function for the hidden layer
- Softmax activation function for the output layer
- Cross-entropy loss function
- Stochastic Gradient Descent (SGD) optimizer
Epoch 1, Accuracy: 96.12%, Avg Loss: 0.2188
Epoch 2, Accuracy: 96.98%, Avg Loss: 0.0875
Epoch 3, Accuracy: 97.41%, Avg Loss: 0.0561
Epoch 4, Accuracy: 97.63%, Avg Loss: 0.0383
Epoch 5, Accuracy: 97.63%, Avg Loss: 0.0270
Epoch 6, Accuracy: 97.69%, Avg Loss: 0.0193
Epoch 7, Accuracy: 97.98%, Avg Loss: 0.0143
Epoch 8, Accuracy: 98.03%, Avg Loss: 0.0117
Epoch 9, Accuracy: 98.03%, Avg Loss: 0.0103
Epoch 10, Accuracy: 98.06%, Avg Loss: 0.0094
Epoch 11, Accuracy: 98.06%, Avg Loss: 0.0087
Epoch 12, Accuracy: 98.16%, Avg Loss: 0.0081
Epoch 13, Accuracy: 98.16%, Avg Loss: 0.0078
Epoch 14, Accuracy: 98.18%, Avg Loss: 0.0075
Epoch 15, Accuracy: 98.19%, Avg Loss: 0.0074
Epoch 16, Accuracy: 98.20%, Avg Loss: 0.0072
Epoch 17, Accuracy: 98.24%, Avg Loss: 0.0070
Epoch 18, Accuracy: 98.23%, Avg Loss: 0.0069
Epoch 19, Accuracy: 98.23%, Avg Loss: 0.0069
Epoch 20, Accuracy: 98.22%, Avg Loss: 0.0068- GCC compiler
- MNIST dataset files:
train-images.idx3-ubytetrain-labels.idx1-ubyte
gcc -o nn nn.c -lm-
Place the MNIST dataset files in the
data/directory. -
Compile the program.
-
Run the executable:
./nn
The program will train the neural network on the MNIST dataset and output the accuracy and average loss for each epoch.
You can adjust the following parameters in nn.c:
HIDDEN_SIZE: Number of neurons in the hidden layerLEARNING_RATE: Learning rate for SGDEPOCHS: Number of training epochsBATCH_SIZE: Mini-batch size for trainingTRAIN_SPLIT: Proportion of data used for training (the rest is used for testing)
This project is open-source and available under the MIT License.