A Hierarchical Approach for Handwritten Digit Recognition Using Sparse Autoencoder

Higher level features learning algorithms have been applied on handwritten digit recognition and got more promising results than just using raw intensity values with classification algorithms. However, the approaches of these algorithms still not take the advantage of specific characteristics of data. We propose a new method to learn higher level features from specific characteristics of data using sparse autoencoder. The main key of our appoarch is to divide the handwritten digits into subsets corresponding to specific characteristics. The experimental results show that the proposed method achieves lower error rates and time complexity than the original approach of sparse autoencoder. The results also show that the more correlated characteristics we define, the better higher level features we learn.