Preface

getting started

Before getting started my journey to artificial intelligence:

This story starts from python and google colab. I probably will write a prequal somday. Hopefully, readers have an experience on python and colab.

There are two strong candidates for framework, I'd like to call, for neural network simulation or explorer. One is Tensorflow, google is said to support, and the other is pytorch, facebook is said to support. All of them is an open source, and draws a large attention from a.i. community. There is the DirectML by microsoft as well. I rarely use it.

Tensorflow has a big transition from version 1.x to 2.x. I'm not a big fan of tensorflow so far, but the coding style of tensorflow seems hard to read. In the meantime, Keras is regarded as a frontend framework for tensorflow. Keras helps to make codes simple to read. Hence, I will use Keras to use tensorflow.

Pytorch is a kind of extened version of numpy. If a reader is familiar with numpy, a powerful library for matrix and array in python, he/she can easily read codes here.

There are pros and cons for both sides. Considering GPU support for each framework, tensorflow seems winning at 2021. Pytorch's AMD gpu support looks insufficient; only a few AMD gpus can be utilized for both frameworks. Contrarily, NVIDIA gpus have a large avalilable models for both.

Let's get started!

This is not an structured tutorial for Keras and Pytorch. Who need tutorials is recommended to visit an official tutorial page per a framework.