Awesome Open Source
Awesome Open Source

Build Status MIT License

Cranium is a portable, header-only, feedforward artificial neural network library written in vanilla C99.

It supports fully-connected networks of arbitrary depth and structure, and should be reasonably fast as it uses a matrix-based approach to calculations. It is particularly suitable for low-resource machines or environments in which additional dependencies cannot be installed.

Cranium supports CBLAS integration. Simply uncomment line 7 in matrix.h to enable the BLAS sgemm function for fast matrix multiplication.

Check out the detailed documentation here for information on individual structures and functions.


  • Activation functions
    • sigmoid
    • ReLU
    • tanh
    • softmax (classification)
    • linear (regression)
  • Loss functions
    • Cross-entropy loss (classification)
    • Mean squared error (regression)
  • Optimization algorithms
    • Batch Gradient Descent
    • Stochastic Gradient Descent
    • Mini-Batch Stochastic Gradient Descent
  • L2 Regularization
  • Learning rate annealing
  • Simple momentum
  • Fan-in weight initialization
  • CBLAS support for fast matrix multiplication
  • Serializable networks


Since Cranium is header-only, simply copy the src directory into your project, and #include "src/cranium.h" to begin using it.

Its only required compiler dependency is from the <math.h> header, so compile with -lm.

If you are using CBLAS, you will also need to compile with -lcblas and include, via -I, the path to wherever your particular machine's BLAS implementation is. Common ones include OpenBLAS and ATLAS.

It has been tested to work perfectly fine with any level of gcc optimization, so feel free to use them.


#include "cranium.h"

This basic example program is the skeleton of a classification problem.
The training data should be in matrix form, where each row is a data point, and
    each column is a feature. 
The training classes should be in matrix form, where the ith row corresponds to
    the ith training example, and each column is a 1 if it is of that class, and
    0 otherwise. Each example may only be of 1 class.

// create training data and target values (data collection not shown)
int rows, features, classes;
float** training;
float** classes;

// create datasets to hold the data
DataSet* trainingData = createDataSet(rows, features, training);
DataSet* trainingClasses = createDataSet(rows, classes, classes);

// create network with 2 input neurons, 1 hidden layer with sigmoid
// activation function and 5 neurons, and 2 output neurons with softmax 
// activation function
size_t hiddenSize[] = {5};
Activation hiddenActivation[] = {sigmoid};
Network* net = createNetwork(2, 1, hiddenSize, hiddenActivation, 2, softmax);

// train network with cross-entropy loss using Mini-Batch SGD
ParameterSet params; = net; = trainingData;
params.classes = trainingClasses;
params.lossFunction = CROSS_ENTROPY_LOSS;
params.batchSize = 20;
params.learningRate = .01;
params.searchTime = 5000;
params.regularizationStrength = .001;
params.momentumFactor = .9;
params.maxIters = 10000;
params.shuffle = 1;
params.verbose = 1;

// test accuracy of network after training
printf("Accuracy is %f\n", accuracy(net, trainingData, trainingClasses));

// get network's predictions on input data after training
forwardPass(net, trainingData);
int* predictions = predict(net);

// save network to a file
saveNetwork(net, "network");

// free network and data

// load previous network from file
Network* previousNet = readNetwork("network");

Building and Testing

To run tests, look in the tests folder.

The Makefile has commands to run each batch of unit tests, or all of them at once.


Feel free to send a pull request if you want to add any features or if you find a bug.

Check the issues tab for some potential things to do.

Get A Weekly Email With Trending Projects For These Topics
No Spam. Unsubscribe easily at any time.
C (274,060
Machine Learning (30,085
Neural Network (8,403
Classification (3,032
Embedded (1,783
Travis Ci (1,708
Continuous Integration (1,576
Regression (1,286
Matrix (1,208
Header Only (784
Artificial Neural Networks (709
Portable (502
Blas (269
C99 (260
Related Projects