Pattern Recognition using neural networks

Question

I want to design a neural network which can be used for pattern recognition instead of traditional classification. Is it possible for me to build such a neural net which can generate a pattern as an output ? Also, how to deal with text data sets in case of neural networks ?

Answer 1

your question is not a specific one ,so answer to your question could be a book, anyway yes you can build a network that can generate a pattern

if you want to generate pattern in output something like your input data , you can use auto-encoder type architecture this type of architecture try to build an output like the one placed in input ,so to train these type of networks your loss function is the difference/distance between generated pattern and input pattern , and the network try to minimize this loss

if your desired output is different from your input you still can use auto encoder you train the network in the same way and put desired output in between then you can get what you want

there is a whole bunch of these architecture out there like GAN , any generative model would serve your cause , for beginning you can start with Haykin's book or this great website

and about text and how to generate text pattern first you have to choose representation for your input text like embedding matrix then you feed it to your network and train it , for text generation sequential model (like LSTM,..) is commonly used