NELSONZHAO/zhihu

Implements 14+ end-to-end deep learning projects using TensorFlow and Keras, spanning character-level text generation with LSTMs, Word2Vec embeddings, sequence-to-sequence models with attention mechanisms, image generation via GANs and style transfer, and CTR prediction models (DeepFM, xDeepFM, AutoInt). Each module includes practical implementations on standard datasets (MNIST, CIFAR, Zhihu translation corpus) with comparative analysis of architectures—such as DNN/LSTM/CNN for sentiment analysis and batch normalization ablations—bridging theory with reproducible code examples.

3,529 stars. No commits in the last 6 months.