Category: 2020 Fall

Code: https://github.com/vionwinnie/pyspark-horse-race-predict 一般的機器學習課程只介紹怎樣用 JUPYTER NOTEBOOK 用 CSV FILE 去訓練 MACHINE LEARNING MODEL。但實際上在商業情況下，一>旦把模型 DEPLOY ，做 SCORING 的時候就要將 BIGDATA 既 DATAPIPELINE 連接到去 MODEL。我想介紹怎樣用 PYSPARK 建立 ETL PIPIELINE ， 從 HIVE DATABASE 抓取數據及整理數據 再傳送到 tensorflow 模型去計算結果並儲存數據。 這是一個 ONLINE MACHINE LEARNING SYSTEM 和 BIG DATA 應用的簡介。(Spark, neo4j, Python, Tensorflow, ETL, end-to-end machine learning model deployment) Language: CantoneseDate & Time:…

FastAPI is becoming a popular library in building REST API endpoint.  The documentation interface and typing conversion makes it much easier and faster to develop and debug when compared to vanilla Flask.  I am going to share some experience in migrating from Flask to FastAPI, as well as integrating authentication and testing with FastAPI. Language:…

There are many GUI toolkits in the standard python but when it comes to embedded device applications, graphical user interfaces must run on Liquid Crystal Displays(LCDs).In developing GUIs for this systems, one needs to go into a much lower level language paradigm using micropython.This talk will discuss the general techniques on developing GUIs in LCDs…

JAX is a library similar to NumPy, redesigned by Google for GPUs and TPUs. A wave of new ML frameworks, based on JAX, are looking for a chance to become an alternative to PyTorch and TensorFlow. In this session, I try out three JAX frameworks (Objax, Flax, and Elegy), showing the steps to solve Kaggle’s…

This is a sharing session on how the speaker had built a production backend for the website that served PyCon APAC 2020 within 4 days. During these 4 days, the speaker had to plan the structure of webapp that interacted with both front-end and back-end, together with a middle-ware that used to check the validity…

Slides: https://docs.google.com/presentation/d/14yapVXWpFHqgBEiToD7Qkp1jNG0ksE6MnELWxGEfVP4/edit?usp=sharing Real Time Monitoring for Machine Learning in Production “As the number of production machine learning use-cases increase, we find ourselves facing new and bigger challenges where more is at stake. Because of this, it’s critical to identify the key areas to focus our efforts, so we can ensure our machine learning pipelines are…

Time series data without missing values or gaps are a general prerequisite in performing analyses.but what can we do when our data contains gaps and what techniques can we use to fill thesevalues? Let us revise some of the widest used gap-filling techniquesSome of the techniques I will cover in this talk are: Linear interpolation…

Oftentimes, company websites are thought of as a single monolithic application. However, as many of you know, websites often host many different applications through a single DNS. We’ll do a dive into how Clover balances the intricacies of allowing independent and customizable content management along with complex applications in one single platform using the power…

If you are anything like me, when you started with Python ‘print’ was the debugger of choice. But you likely found that was slow, tedious, and didn’t cut it for more complex problems. Let’s dive into methods for debugging remote python in environments such as Raspberry Pi, Docker containers, remote Linux Servers, and Jupyter Notebooks…

In this session, I will share about 2 topics: Developing cloud-native application development using Python with MySQL Database Service Implementing agile DevOps with MySQL docker containers on Kubernetes” Remark:I am providing first-hand information on our newly launched MySQL Database Service on OracleCloud (https://www.mysql.com/cloud/) as well as MySQL deployment on Kubernetes. I believethese topics will benefit…