MNIST Handwriting Recongnition Neural Networks. Feed-Forward back propagation, k-nearest-Neighbor, and Radial Basis Function
Hand Writing Recognition and Simple CAPTCHA Neural Networks: Feed-Forward Backpropagation network, K-nearest Neighbor network, Radial Basis Function Network
Min "Ivy" Xing, Zackery Leman
These networks work by reading in an image (image data provided by MNIST) and then selecting the number 0-9.
All of the networks are objects that need to be created via a constructor in another file (testAllNetworksQueue). If the testAllNetworksQueue.java file is compiled and run, then for demonstration purposes, one object of each neural net type will be created and run. (Default w/ 8 avalible core assumption). For this to work make sure you first do the following:
- First unzip the NeuralNetOutput.zip in a directory.
- Change all instances of "/Users/zackeryleman/Desktop" in testAllNetworksQueue.java to be the file path of the NeuralNetOutput folder.
// NOTES: To run these networks we had to pass the argument " -Xmx3000M" to the java virtual machine to increase heap memory allocation. // This code requires that the newest Java Runtime environment is installed. Fatal bugs in older runtimes will prevent the networks from functioning. //When testing Captchas if you get a, cannot find file error, either add or remove the "src/" before all file paths in the CAPTCHA loading method.