I’m start reading this book <Deep Learning with TensorFlow 2 and Keras> those days, and will keep posting what I learnt from the book here.

Google Colab is one of the best place to start your Machine Learning. Sometime you may want to upload images to the notebooks from your local. Fortunately you can easily make it done throught the built-in API.

Python introduced async/await syntax from Python3.5. it makes your code non-blocking and speedy. Developers can use it to build a high-performance / NIO web services like NodeJS. Most of the Python web developers are familiar with Flask. But unfortunately flask has no plan to support the async request headers. Sanic is a Flask-like webserver that’s written to go fast. It was inspired by uvloop.

A trained convolutional layer is made up of many feature detectors, called filters, which slide over an input image tensor as a moving window. This is a very powerful technique and it possesses several advantages over the flatten and classify method or deep learning.

One of the biggest headaches of using deep neural networks is that they have tons of hyperparameters that should be optimized so that the network performs optimally. Below are some notes coming from Deep Learning Quick Reference.

After reading several books about deep learning, now I can use keras / tensorflow to train some models, but the mathmatical implementations behind the libraries are still have to follow.

This is the notes when I read the book Scaling Your Nodej.js Apps.

Layered Architecture

It is based on the logistic seperation of concerns of your application (or platform) into layers. And the layers must comply with the following points:

This is the notes when I read the book Scaling Your Nodej.js Apps.

An increasing in incoming traffic could affect your system in different ways; we can describe these as direct or indirect.

聚类是接触无监督式学习时常见的一个问题。通常在你想将一个大的数据集重新编排组织成小的群组，或者是将其按逻辑相似性分解成小的群组时我们会考虑聚类。例如，你可以尝试将人口普查数据中的家庭收入分成三个群组：低收入，中等收入及高收入。如果你将家庭收的数据集做为指定的一种聚类算法的输入，你将期望得到三个数据聚点 及其对应的前述三个收入分组的平均值做为结果。即使是这种家庭收入数据的一维聚类问题，你也可能很难通过手动解决——难点在于无法快速地确定两个分组之间的边界值。你可能可以参考政府部门所定义的收入分组的标准，但无法确定这类定义的标准是几何均衡的; 这类的标准可能是被政策制定者所发明出来的，因此它可能无法精确地表达数据之间的正真关系。

If you enable HTTPS and set up the certifications correctly, which means data will not be decrypted or modified during the transportation. Today I try to enable SSL to my website. Here is what I did to make it happen: