Life can only be understood looking backward. It must be lived forward.

— The Curious Case of Benjamin Button

This is my second article (first on Convolution Neural Network) of the series on Deep Learning and Reinforcement Learning.

There are many sequential modelling problems in day-to-day life: machine translation, voice recognition, text classification, DNA sequencing analysis, etc. Most of these problems belong to supervised learning. Either the input or output of the models is sequential with variable size. Normal neural networks have difficulties dealing with input and output of varied sizes.

In the following sections, I am going to explain…


With a Simple Implementation of ResNet in Keras

Learn from yesterday, live for today.

— Albert Einstein

Historically Machine Learning on images had been a big challenge until 2012, when AlexNet won the first prize of the ImageNet competition and got the attention of the general public. Ever since then, Convolution Neural Network or CNN has evolved so fast that AlexNet has soon become a thing of the past. CNN has surpassed human performance significantly in terms of image processing.

However, the concept of CNN was already brought out and put into practice more than 30 years ago. The main reason why CNN or deep-learning has become so…


Hands-on Tutorials

An Example of Microkernel Architecture with Python Metaclass

What is Microkernel Architecture

The Microkernel Architecture is sometimes referred as Plug-in Architecture. It consists of a Core System and Plug-in components.

Th Core System contains the minimal functionality required to make the system operational. The Plug-in components are standalone and independent from one another, each of which enhances or extends the Core System with additional capabilities.

The Core System also maintains a plug-in registry, which defines information and contracts between the Core System and each Plug-in component, such as input/output signature and format, communication protocol, etc.

Image from O’Reilly: Software Architecture Pattern

The pros and cons of the Microkernel Architecture is quite obvious. On one hand, the separation of…


Distributed Training in Kubernetes Cluster with a Simple YAML File

A few days ago I was exploring the RandomForest model with a huge dataset. To my disappointment, the training data couldn’t fit into the memory of the single working node. I had to manually split the training data to train a few RandomForest models, before combining the models into a single one.

This prompted me to an idea: can I come up with a framework to automate and distribute the trainings for ensemble-able models with huge datasets?


Common problems of machine learning system and sharing of personal experience

With the democratization of Machine Learning (ML) technologies and the advancement of ML tooling and frameworks, developing and deploying ML systems has become fast and cheap. However, maintaining such systems over time proves to be difficult and expensive, due to the peculiarity and unique characteristics of ML systems.

I am going to briefly summarize some of the technical debts of ML systems from 3 perspectives, which are Data Dependency, ML System Peculiarity and System Architecture and Design respectively. After that, I will try to relate some of them to a ML project that I have done previously.

ML System Technical Debts

Due to Data Dependency

ML System’s Data…


Why We Need Machine Learning Platform and What it Consists of

Why We Need a Machine Learning Platform

Machine Learning Models Become Competitive Business Advantages

Business competes on differences and unique competitive advantages. The democratization of AI technologies has also enabled more and more businesses to apply their proprietary data to solve their proprietary business problems through the creation of proprietary models. The benefits of such a proprietary model-driven approach could be huge, one of the famous examples is Netflix Recommendation Model save it $1 Billion dollars yearly.

As models become organization’s competitive business advantage, the ability to develop, deploy and integrate them into business operations in a scalable and efficient way becomes very important.

Operationalizing Machine Learning is not Easy

However, Operationalizing machine learning models is not just wrapping model…


I just got the certification for AWS Certified Machine Learning - Specialty last week. During the preparation, I was impressed by both the extent of toolings that AWS offers and how they could reduce the undifferentiated heavy-lifting for all phases of machine learning lifecycle. Therefore, I decided to write an article to explain in more details on this perspective.

I will start with an overview of machine learning processes in the real-world, followed by an brief introduction of AWS machine learning stack. Then I will illustrate it through a concrete use case of machine learning project.

Machine Learning in the Real-World is Much More than ML Code

Many of you might…


When given a new data set at the beginning of a Machine Learning project, one of the laborious tasks is to explore various ways of Feature Engineering as well as to experiment with different Models and Hyperparameters, in order to find out the best combination of Feature Engineering, Model and Hyperparameters that works best.

A good baseline is essential for Data Scientist to bring the project forward. However, baseline exploration is often time-consuming. It is an undifferentiated heavy lifting that should be taken away from data scientists.

Therefore, one question that often pops up in my mind is that: Given…


Machine Learning Platform Accelerate Application of Machine Learning Models

With the rapid development in the fields of modelling and algorithm of machine learning in the past decade, more and more attention have been oriented towards the application and adoption of enterprise machine learning. Some of the questions include: How do we make use of the latest data science models? How do we reduce the time-to-market of machine learning models? How do we seamlessly integrate machine learning into business and operational processes?

Machine learning platform seems to be the answer. A machine learning platform is a set of tools and technologies to support and automate machine learning workflow and lifecycle…

Chen Yanhui

Machine Learning Engineer

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store