After talking about “Machine Learning” in my previous article, this time I want to talk about “Deep Learning” which is a hot topic in AI world in recent years. Deep learning is a subset of Neural Networks and Neural Network itself is a subset of Machine Learning. In below diagram, you can find how this topics are related to each other.
Actually a Deep Learning network is a Neural Network which has more than 1 hidden layer. In below picture, there is a simple comparison between Deep Learning network and Neural Network.
Deep learning was introduced in almost 2006, but the concept of neural network is much older and refers to 1970’s. The golden age of neural network was 1980’s and many inventions like autonomous car or digit recognition was created at that time. But it was out of focus for a while. A reason for that, is Neural Network needs a lot of computational work and the computers at that time were not very high speed processing. In recent decades, by advances in computer process units, deep learning and neural network are back to business.
Previously, researchers for simulation purposes, just used some ready toolboxes for neural network in software like Matlab. But in these years, many framework has developed for Machine Learning and Deep Learning. For example “Keras”, “TensorFlow” and “PyTorch” are famous frameworks of Python language and “Caffe” is famous deep learning framework of C++ language. Some Frameworks, like PyTorch, due to huge computational works of deep learning models, only run on GPU which is somehow faster than working with CPU.
Deep Learning made some advances in Computer Vision Problems which will be talked in next article.