This project contains Machine Learning basics wrote by myself
About Input
- The input of each model is a double matrix
- If using batch GD, please set the input as the whole training data
- If using SGD, please set the input as one training instance per time
- If using SGD with mini-batch, please set the input as one batch per time
Models
- Logistic Regression
- Multi-class Logistic Regression (2-class also supported)
- Written in C++ 11
- Hiddenlayer
- Hiddenlayer for mlp (feedforward neural network)
- could be used anywhere you want
- activate function includes:
- tanh (default)
- sigmoid (update fuc need to be changed)
- ReLU (update fuc need to be changed)
- Multi-layer Perceptron (Feedforward NN)
- Single hidden layer MLP
- mini-batch supported
- Retricted Boltzmann Machine
- First commit, bug fixing
- RBM done
- Autoencoder
- done
- CNN
- Convolutional layer
- need to construct a complete CNN
- need to complete the update function
- the most important step is to calculate the gradient of each layer
- RNN
- Recurrent NN with word embeddings
- An example. mini-batch is all words in a sentence represented with word embeddings