Skip to content

Latest commit

 

History

History
70 lines (37 loc) · 1.52 KB

File metadata and controls

70 lines (37 loc) · 1.52 KB

MachineLearningNotes

My notes on machine learning

Files

MultivariateLinearRegression

MultivariateLinearRegressionUsingGradientDescent.ipynb

MultivariateLinearRegressionUsingNormalEquation.ipynb

TorchAutogardLinearRegression.ipynb

Instead of calculating derivatives myself, we use the torch.autograd to compute gradients.

kNN

kNNbyKDTree.cpp

Implemented k-nearest neighbors algorithm by C++.

NeuralNetworks

BPNetworks.ipynb

Implemented backpropagation algorithm. Trained a 2-layer neural networks.

Its model file is trained.npz.

MultiHiddenLayerNN.ipynb

Implemented a multilayer neural networks, of which activation functions are all sigmoid.

Using PyTorch only for matrix multiplication.

Verified that sigmoid activation function can cause vanishing gradient.

MultiHiddenLayerNNReLU.ipynb

Replaced sigmoid funtions in MultiHiddenLayerNN.ipynb with ReLU except the last layer. Vanishing gradient solved.

Its model file is model_ReLU94.pth,whose test accuracy is 94%.

PyTorchNN

torchNN.ipynb & torchNNretrain.ipynb

Updated from BPNetworks.ipynb.

Using PyTorch to do matrix multiplication to speed up.

Its model file is trained_model_tensor.pt and trained_model_tensor2.pt

NaiveBayes

NaiveBayesClassifier.ipynb

EnsembleLearning

WeightedVote.ipynb

Paddle

image_recognition_paddle.ipynb

text_classification_paddle.ipynb

Its model file is model_final.pdparams.

References

THE MNIST DATABASE of handwritten digits