Dive-Into-Deep-Learning

Concluding 10 coursework and 1 final projects. The course lead us to dive into various aspects of deep learning, including simple Softmax Regression/Classification, Logistic Regression/Classification; Convolutional Neural Networks like LeNet, ResNet and DenseNet; and Recurrent Neural Networks like GRU, LSTM (single layer and multi-layer). The repo includes experiments of constructing neural networks with MXNet (Amazon) on prestigious topics and datasets, like MNIST, CIFAR-10, ImageNet, IMDB dataset, stocks dataset, e.t.c. The projects was experimenting MXNet on Quora language dataset. It could be found on Kaggle: https://www.kaggle.com/c/quora-insincere-questions-classification/discussion.

• Deep Learning tutorial: http://d2l.ai/
• Course github repo: https://github.com/d2l-ai/berkeley-stat-157
  Assignment and corresponding answers are included in the course github repo.

Assignment 1: Familiar with MXNet and NDArray

Assignment 2: Math, Sampling, Numerical Stability

Assignment 3: Logistic Regression

Assignment 4: First Kaggle Competition - House Price Prediction

Assignment 5: Covariate Shift

Assignment 6: Kaggle Competition CIFAR-10

Assignment 7: Kaggle ImageNet

Assignment 8: Language Model - 'Poeting' like Shakespear

Assignment 9: Time Series using RNN with Stock Price Dataset

Assignment 10: Bidirectional Encoder