Abstract
Deep learning has demonstrated tremendous success in variety of applicationdomains in the past few years. This new field of machine learning has beengrowing rapidly and applied in most of the application domains with some newmodalities of applications, which helps to open new opportunity. There aredifferent methods have been proposed on different category of learningapproaches, which includes supervised, semi-supervised and un-supervisedlearning. The experimental results show state-of-the-art performance of deeplearning over traditional machine learning approaches in the field of ImageProcessing, Computer Vision, Speech Recognition, Machine Translation, Art,Medical imaging, Medical information processing, Robotics and control,Bio-informatics, Natural Language Processing (NLP), Cyber security, and manymore. This report presents a brief survey on development of DL approaches,including Deep Neural Network (DNN), Convolutional Neural Network (CNN),Recurrent Neural Network (RNN) including Long Short Term Memory (LSTM) andGated Recurrent Units (GRU), Auto-Encoder (AE), Deep Belief Network (DBN),Generative Adversarial Network (GAN), and Deep Reinforcement Learning (DRL). Inaddition, we have included recent development of proposed advanced variant DLtechniques based on the mentioned DL approaches. Furthermore, DL approacheshave explored and evaluated in different application domains are also includedin this survey. We have also comprised recently developed frameworks, SDKs, andbenchmark datasets that are used for implementing and evaluating deep learningapproaches. There are some surveys have published on Deep Learning in NeuralNetworks [1, 38] and a survey on RL [234]. However, those papers have notdiscussed the individual advanced techniques for training large scale deeplearning models and the recently developed method of generative models [1].