2025-03-02 –, SS 117
Given the rise of popularity of AI and machine learning, more and more applications are leveraging on these powerful tools. They open up so many possibilities in with proper integration to your system.
There are many aspects of AI and machine learning that can affect its performance such as quality of data, type of algorithm, structure of the models, and deployment environment. However, one important aspect is often overlooked and frowned upon.
<u>Model maintenance and validation</u> is often left on the sidelines as users of AI and ML often focus on the other aspects. Improving your data for example is a good way to increase model performance - on paper. But same as any other software application, any kind of ML model needs proper maintenance to make sure that its result is still valid and accurate over time.
This session will help us understand how to implement a process of validating and maintaining models using python. We will deal with concepts such as data drift and model drift. We will also implement and experiment on different validation approach.
This workshop is designed to highlight the following concepts:
- How to measure the performance of a machine learning model
- What is drift? How does this affect the overall usability and reliability of your ML applications
- How to correct problems with model performance using existing python tools and libraries
What do I need to know before attending the session (this is an intermediate level session):
- have implemented at least one ML model (either thru projects or personal hobby)
- basic statistics (standard deviation, mean, median)
- any kind of software development experience (junior - senior)
This workshop is designed for two hours with the following breakdown
- Quick session on machine learning applications life cycle (20 mins)
- Hands on experiment on how a model performance degrades over time (30 mins)
- Quick discussion on how to mitigate this problem (10 mins)
- Hands on implementation of model validation and drift detection (40 mins)
- Wrap up discussion and additional resources (10 mins)
At the end of this session, you are expected to gain the following:
- Ability to identify problems with model performance
- Understanding of how to correct model performance and apply them in your projects
- Proper tools and libraries that you can use
No pre-setup is needed as we will just use jupyter notebooks. For those interested, I will include a fast-api based implementation that you can try on your own time.
Intermediate
Category:Machine Learning and Artificial Intelligence
Workshop Duration:2 hours
Ninz is a machine learning engineer at an agri-tech company, with nearly 13 years of experience in the tech industry, specializing in machine learning, data science, and research. Their work primarily focuses on solving challenges in climate, agriculture, and biodiversity.
Beyond their professional role, Ninz actively supports tech and research initiatives behind the scenes. Whether working on projects over a good cup of coffee or riding trails high in the mountains, they are always engaged in innovation and exploration.
And above all, they have a deep love for custard buns.