• This is a complete report that explains how the neural network works from the very scratch.
  • This is a comprehensive neural network tutorial that covers all aspects of the topic.
  • You only need to be familiar with +-x/ and derivative calculation.
  • We will go through:
    • Forward propagation process with activation function (sigmoid and softmax) in the hidden layer and in the output layer.
    • Backward propagation process with derivative calculation and cross-entropy of each parameter (weight and bias).
    • Update each parameter (weight and bias) simultaneously.
• This is a classification example.
  • Dataset: Iris dataset with 4 features and 3 classifications
  • Implement
    • data pre-processing (one-hot encode),
    • forward propagation,
    • backward propagation,
    • activation functions,
    • cross-entropy (loss calculation),
    • derivative calculation,
    • update parameters (weight and bias) simultaneously,
    • training model on the train dataset,
    • prediction in the test dataset,
    • accuracy calculation on the test dataset
  • from the scratch only via Numpy and Pandas.

• Digital Image Pre-Processing and pre-knowlegde
  • Bit depth and Gray Levels (the specific value of each pixel)
  • skimage.io.imread read pictures as y,x,c where c is RGB 3 channels.

• Convolution Neural Nework in One Image.

• Principle Component Analysis (PCA) from the scratch and via Sklearn

• Medical Image Classification (Normal vs Pneumonia)

• Iris-Eye Image Segmentation via U-net in Keras

• NLP supervised classify topic of documents

• Twitter sentiment analysis using Word2vec bi-LSTM

• Metrics for ML

Math & Plot tips

interesting notes

1. https://youtu.be/HGwBXDKFk9I
2. https://youtu.be/NItHNRc3awY
3. https://github.com/ikatyang/emoji-cheat-sheet/blob/master/README.md