This is HW1 repo for Practical-Introduction-to-Neural-Network. It aims at implementing basic fully connected neural network and perceptron using Tensorflow. Two datasets are included, separately MNIST for handwritten digits, and Iris flower dataset. The neural network implements some basic optimization methods, including Batch GD, Adam optimizer. And different initialization approaches are investigated, like random normal, Xvaier and He initialization. The goal of this homework is to get familiar with tensorflow basic implementations on neural networks, which paves the way ahead for the more complicated neural networks.