Artificial Neural Network

  Self Organizing Map Ideas first introduced by C. von der Malsburg (1973) and developed and refined by T. Kohonen (1982)  It is a type of Unsupervised competitive learning. It is  primarily used for the organization and visualization of complex data. SOM Architecture                         It is two layers of Neuron Architecture 1. Input layer 2.Output Layer. Each input Neuron is                          connected to each output Neuron.It is fully connected Network. The output  map usually has                          two dimensions.   O ne and three dimensions are also used    Neurons in output map can be laid out in different patterns o    Rectangular o    Hexagonal o    Random  SOM Training         “Neighbourhood” is a n important concept in SOM training.             The output map neurons that adjoin the winner           Neighborhood size describes the distance between neighbors and  the winning neuron.            Neighbors weights are also modified Neighb

  WHY DEEP NEURAL NETWORK : Difference between neural network and Deep neural network .  Labeled data set is the prime requirement of Supervised learning . As a .  massive data is generated every day, AI Projects are innovated with rapid speed and choice of the algorithm is a critical issue: Choice of the algorithm depends  1.       Number of data sets (samples) and attributes  in that data set.(Dimensions of input and                 output). 2 .      Linearity and non-linearity in that data set 3   Performance. Neural Network or Deep Learning:  Algorithmic steps of neural network are   1.    Need to decide Architecture of NN i.e.Number of hidden layers and number of units in that hidden layer. 2.     Apply training algorithm by Deciding hyperparameters like, Transfer function, No of iteration, learning rate etc to train the network. 3.     Test the performance   Major disadvantages are Ø   If required performance is not achieved we need to repeat step 1 to 3 again.This

Deep Learning: Convolution Neural Network:  A convolution  neural network is a type of supervised learning. Convolutional Neural Network strength: •        Feature extraction and classification are integrated into one structure •          Fully adaptive. •          Network extracts 2-D image features at increasing dyadic scales. •        Relatively invariant to geometric, local distortions in the image. •        Applications: Hand-written digit recognition, face detection, and face recognition   A Convolutional neural networks are designed to process two-dimensional (2-D) image A CNN consists of three layers   (i) Convolution layers (ii) Sub-sampling layers (III) Outputlayer or fully -connected layers(fc). Network architecture is shown in fig1.                                                         Figure1:Network Architecture Network layers are arranged in a feed-forward structure: Each convolution layer is followed by a sub-sampling layer and the last conv