Robust-first Computing Wiki

• Create a trainable neural network based on input neurons, unknown neurons, and output neurons. Once trained this should be able to decently classify objects that it was trained to recognize in input images. Create a visual representation of how the neural network feed weights into the next row of neurons, then will back propagate based on the error calculated.

• Create the three different types of neurons I specified above, that will be inherited from a parent generic neuron quark
• Create some sort of connection/communication system where a all neurons in one row can transmit/talk data to any neuron in the rows adjacent to them no matter how big the rows get
• Figure out how to feed images into a ulam program and break them down into its byte representation, and then feed those to the input neurons
• Find a simple training data-set to feed into this neural network
• Try different images after the network has been trained

Week 7 Update:

• Now understand the concept of CNN better through messing with tiny_dnn
• Found “final” design for blocks of CNN
• Will now work on communication between the blocks in the CNN
• Also will work on fixing a size for the CNN based around a smallish image