Recognition of Handwritten Digits Using DNN, CNN, and RNN

Deep learning is the domain of machine learning that implements deep neural architectures, with multiple hidden layers to mimic the functions of the human brain. The network learns from multiple levels of representation and accordingly responds to different levels of abstraction, where each layer learns different patterns. Handwritten digit recognition is a classic machine learning problem to evaluate the performance of classification algorithms. This paper focuses on the implementation of deep neural networks and deep learning algorithms. The NN algorithms such as DNN, CNN, and RNN are implemented for the classification of handwritten digits. The algorithms are implemented on various deep learning frameworks and the performance is evaluated in terms of accuracy of the models. The best accuracy is of CNN 99.6% model and the error rate of algorithms ranges from 0.2–3%.