Use TensorFlow.js to do Kaggle -- A practice of Deep Learning Training

In a word, the mnist-10 dataset is for hand-written digits recognition, where each image is a black-white image with a size of 28 * 28. Let's have a look at the train.csv This code will take one data from the dataset and print it out. As we can see from the console, the dataset has two parts xs which is a hashmap(dictionary) of each pixel and its corresponding value (ranges from 0–255) and ys which is the corresponding label. Okay, firstly, we need to wrap xs and ys as tensors so that the TensorFlow.js can smoothly deal with them.