It is my project based on Python 3, Tensorflow ans Scikit-learn to the course Hybrid Systems of Center of Informatics of UFPE (2016.1). This course is related with my Master Degree in Deep Learning. The idea is to extract features from MNIST dataset and then use SVM to classify. We show that improves the accuracy when compared with the regular CNN and is faster than ELM of worst accuracy.

To run it, you shoult type: