NeuralNetworks

Neural network are now used everywhere in all areas of industry especially for deep learning. Mostly used for pattern recognition (analysis of photos to recognize people or faces, recognition of fish swarms in sonar readings, recognition and classification of military vehicles in radar scans, or any number of other applications) Neural Networks can also be trained to recognize spoken language and hand written text or to control self driving cars etc. During this assignment our objective is to gain practical experience with basic aspects of Deep Learning, especially with deep artificial neural network (DNN). Artificial neural networks can be defined as a set of algorithms, modeled loosely after the human brain. They help us ro cluster and classify different types of data set. We will experiment different type of works we can do with DNN and also see with what kind methods and design structures can be used to improve results. DNN requires 3 main characteristics. First of all the choice of the model depending on the data representation. Secondly the learning algorithm and finally the robustness of the DNN. To obtain better results on our different experiments we will make vary hyperparameters (learningRate,batchSize and epochs) but we will also pre-processing the data using different methods and also try different optimisations as the Stochastic Gradient Descent method that could be used for finding a local minimum of a differentiable function.