What is the basic difference between ANN model and Deep Learning? I am not expecting that number of hidden layers are more than ANN. Can anyone have other knowledge please share.
As you mentioned, the difference is not the number of layers only, but in the way, as the layers are combined and its respective roles. For instance, in convolutional networks (a deep learning model) a layer has the goal of modeling a specific pattern (i.e., simple edges/parts of objects), which can be forward to successive layers, while in Radial Basis Function Network (RBF - an ANN model) a layer has a role of projecting the data on a subspace where the problem to be linearly separable.
There is no difference. 'Deep Learning' is the label we came up with to sell neural network models after we improved our techniques to allow us to train neural networks that were far deeper than before. There are four key advances that were made from 1990 to 2012: 1. GPUs deployed to exploit the parallelism for faster computation, 2. Combining convolutional networks with gradient descent, 3. Semi-Supervised learning with Restricted Boltzmann Machines and Autoencoders used to help learn features, and 4. Rectified Linear units that don't have vanishing gradients.
But they are still neural networks and they still use the backpropagation with momentum algorithm developed in 1986 to compute gradients, with some tweaks. Convolutional networks were proposed in a 1980 paper by Fukushima. Whenever you hear 'deep learning' substitute 'state of the art artificial neural network'.