/nn_9869_practice

Primary LanguageJava

9869 Practice: Simple Neural Network

Student Number: 202191382 (zguan#mun.ca)

Sample Output

Question: Can you reuse your synchronization mechanism? Why or why not?

I think yes. On each run of the layer I create a new CountDownLatch and wait forever until all neurons are done. After that the next layer can proceed. Therefore, although we run the same network in many threads at the same time, the synchronization mechanism should still work.

See Concurrent Design part for detailed explanation.

Configurable Items

  • Input Layer Size (Number of neurons, default = 4)
  • Internal Layer Size (Default = 5)
  • Output Layer Size (Default = 3)
  • Activation Function (Defualt = ReLU)

Concurrent Design

Neurons in the same layer can run simultaneously, as they do not interfere each other.

However in the layer level, each neural layer must be run one-by-one. We have a CountDownLatch with capacity = Layer Size for each layer.

Latch Sample

Above is the run method in NeuralLayer.