Implementation-of-MLP-with-Numpy

Implementation of Multi-Layer Perceptron with Numpy and CNN with PyTorch in Image Classification

Image classification

Implementation From Scratch (Numpy)

• Multi-Layer Perceptron (MLP/Neuron Networks) with backpropagation and mini-batch SGD

  Activations:

  • Relu
  • Sigmoid
  • Softmax
  • Leaky_relu

  Layers:

  • Forward
  • Backward
  • Number of layers determined by user

AND

• Pytorch CNN Convoluntional Neuron Network

  Layers:

  • Convoluntional layer (Extract different features from various feature maps)
  • Pooling layer (Max pooling - average pooling: extract dominant features, speed up training speed)
  • Batch Normalization
  • Dropout

Data used:

• CIFAR-10 with 10 classes