Deep Learning

• Framework

TensorFlow	PyTorch
From Google	From Meta (Old: FB / FaceBook)
good at hardware optimization	good for research or study
a set of Keras and with TPU on CoLab	easy to debug and high rate of opensource

• Difference between ML and DL

  Whether the feature extraction/engineering is with classification in the modeling or not
  Does Human expert do the feature engineering or not??

CNN

• a kind of deep learning fields
• it is for image classification.
• Using convolutional calculation, pooling, ReLU in the process

Convolutional Layer

• a convolutional filter or convolutional filters(=kernels) is/are on the input image.
• when this filter moves next or under, it is called as the stride.
• convolutional calculation: multiply the filters of input image with the weight(random value at first)
• the size of activation maps: (input width - filter width) / stride + 1

  channel(depth) of activation maps = how many filters are