xysun / dl-zoo

deep learning code collection

My deep learning code snippets.

Examples:

• Incremental Network Quantization

  • I'm able to achieve 79% test accuracy on CIFAR100 after quantization with the training setup in densenet/quantize.py.

• DenseNet and MobileNetV2

  • Use train.py to train either networks.
  • I'm able to achieve 80.22% test accuracy on CIFAR100 with densenet with the training setup in train.py.

• My solution to XOR parity challenge

  • This is the "warmup" from OpenAI request for research v2

  

• RNN that predicts Shakespeare verse (replicating this tensorflow tutorial with Estimator APIs, also added LSTM as a model choice)

  • Generated Shakepeare after 1000 steps of training on GRU:

    ROMEO:
    No, I shall be the father, that I say,
    The sense of the procled will be the seas.
    
    LADY ANNE:
    What says the seast that we say 'tis a thing and like to see him for the people,
    Which was the seas of t
    

• Demonstrate accelerated learning from batch normalization on mnist: see batch_norm.py

  •