Table of Contents

Source: Deep Learning on Medium


This series of blogs will have an introduction on the deep learning from both theory and implementation aspect.

Before deep dive into deep Neural Network such as ANN, CNN or RNN. Let’s starting our journey from single layer Neural Network. The idea of single layer neural network is first to have weighted linear combination of input variables and then apply activation function to implement non-linear transformation. The difference between logistic regression and perceptron is that, in logistic regression sigmoid function is applied as activation function and in perceptron sign function is applied instead.

For logistic regression, please read my another blog:

For perceptron, hmmmmmmm… I am working on it…

Both logistic regression and perceptron use gradient descent to get parameters. For details of gradient descent, please refer to: