Paper Presentation & Seminar Topics: Digit recognition using neural network

Digit recognition using neural network

Abstract : Seminar ppt The aim of this network is to recognize the numbers one-piece 7-digit LED model. To this end, the network has seven inputs and eleven outputs. Correspond to: 1) LED digit segments is on or off and 2) the outcome of this figure is recognized by the network. Try changing the number by clicking on its components, then test it on the network without training. Most likely be recognized as more digits. The default network is created with 9 nodes in the hidden layer, a rate parameter of 0.5, and a noise parameter of 0.05. If you are really interested in what they mean, soon I'll put a brief discussion of the terms, including a discussion of why the noise is a good thing. (Note - this has been pending for 18 months -. probably will not happen soon) Click on "Create a new network, you can create your own network with the parameters you specify. There are limits on them (noted next to the field) and entry out of bounds will be ignored. These black nodes in the bottom of the input and hidden layer nodes are threshold. They are always at a value of -1 and are used to implement the trigger as a burden on the network . After creating a network by pressing the "Train Network" button, doing this is education and training. And then by the numbers 0 to 9 for the network, test them and compare the actual production of the desired output. The desired output of any number is exactly one node, which corresponds to the value of that number, is on and all the others are off. Any errors that cause the network to provide the weights between the nodes is the number of weight proporational nodes, value, and the derivative of the weight of the production. This continues until all 10 digits, error between actual and expected is very low. At this point the network is trained. Try to give the network a certain income is not a figure and see what has been recorded (if any). This gives an idea of what characteristics of the network must learn to recognize.