viktorc/c-attl3

A C++ deep learning library for the construction and optimization of neural networks ranging from simple feedforward architectures to state-of-the-art convolutional ResNets and LSTMs.

c-attl3's Issues

General TODO

• Test automation using Google Test
• CMake build
• Deconvolutional layer
• Convolution, deconvolution, and pooling for 1st and 2nd rank data
• Broadcast and reshaping layers
• Pixel shuffle layer
• FFT and/or Winograd filtering for CPU convolution
• More comprehensive GPU support and cuDNN utilization
• Passing data between device layers without copying to and from the host
• Network serialization and de-serialization
• PPM codec
• GIF codec
• WAV codec
• IMDB data set provider
• AN4 data set provider
• Working GAN example
• NLP example (word2vec, sentiment analysis)
• Text generation example
• Temporal convolutional network example
• Speech recognition example
• Speech synthesis example
• TF-IDF preprocessor
• LDA preprocessor
• AMSGrad optimizer
• GRU network
• CTC loss function
• Object detection support (YOLO)
• Hyper parameter tuning framework (grid search, gaussian optimization, evolutionary algorithms)

std::thread instead of pthread

In this project pthread is used. This can be migrated to std::thread.