PinnedMember-onlyAbout MeWho I am and What Motivates me to Write — Hello! I’m Khuyen Tran. I have been writing on Medium since December 2019, but I haven’t properly introduced myself so I wrote this article to do so. I major in statistics, but I love playing with data science and Python tools and share them with others in my free time…About Me4 min readAbout Me4 min read
Published in Towards Data Science·Mar 11Member-onlyWrite Readable Tests for Your Machine Learning Models with BehaveUse natural language to test the behavior of your ML models — Motivation Imagine you create an ML model to predict customer sentiment based on reviews. Upon deploying it, you realize that the model incorrectly labels certain positive reviews as negative when they’re rephrased using negative words. This is just one example of how an extremely accurate ML model can fail without…Python9 min readPython9 min read
Published in Towards Data Science·Jan 15Member-onlyHow to Structure an ML Project for Reproducibility and MaintainabilityStart Your Next ML Project With This Template — Motivation Getting started is often the most challenging part when building ML projects. How should you structure your repository? Which standards should you follow? Will your teammates be able to reproduce the results of your experimentations? Instead of trying to find an ideal repository structure, wouldn’t it be nice to have…Data Science7 min readData Science7 min read
Published in Towards Data Science·Jan 1Member-onlyHuman-Learn: Rule-Based Learning as an Alternative to Machine LearningIncorporate Domain Knowledge into Your Model with Rule-Based Learning — Motivation You are given a labeled dataset and assigned to predict a new one. What would you do? The first approach that you probably try is to train a machine learning model to find rules for labeling new data. This is convenient, but it is challenging to…Machine Learning7 min readMachine Learning7 min read
Published in Towards Data Science·Dec 21, 2022Member-onlyBuild a Full-Stack ML Application With Pydantic And PrefectCreate a UI for ML Feature Engineering in One Line of Code — Motivation As a data scientist, you might frequently adjust your feature engineering process and tune your machine learning models to get a good result. Instead of digging into your code to change function parameters: …, wouldn’t it be nice if you could change the parameter values from the UI?Python8 min readPython8 min read
Published in Towards Data Science·Dec 6, 2022Member-onlyRiver: Online Machine Learning in PythonA Fast and Cheap Approach to Update an ML Model in Production — Problem with Batch Learning It is common for data practitioners to use batch learning to learn from data. Batch learning is the training of ML models in batch. An ML pipeline with batch learning typically includes: Splitting the data into train and test sets Fitting a model to the train set Computing the…Machine Learning9 min readMachine Learning9 min read
Published in Towards Data Science·Nov 29, 2022Create Observable and Reproducible Notebooks with HexHow to integrate notebooks into your data pipeline — Motivation Jupyter Notebook is a popular tool for data scientists to explore and process data because of its ease to check code outputs. However, there are several drawbacks to Jupyter Notebook, including: Interpretability issue: As your code gets bigger, the relationship between cells becomes increasingly complex. …Python7 min readPython7 min read
Published in Towards Data Science·Nov 1, 2022Member-onlyDVC + GitHub Actions: Automatically Rerun Modified Components of a PipelineA Perfect Combo to Quickly Iterate on Your DS Project — Motivation Imagine your data pipeline looks similar to the graph below. The pink box represents a stage, which is an individual data process. Dependencies are the files that a stage depends on, such as parameters, Python scripts, or input data. Now imagine Dependencies 2 changes. …Data Science6 min readData Science6 min read
Published in Towards Data Science·Oct 13, 2022Member-onlyCreate a Maintainable Data Pipeline with Prefect and DVCMake Your Pipelines Easier to Support and Maintain — Motivation In engineering, maintainability is the ease with which a product can be maintained to: Meet new requirements Cope with a changing environment Improve the performance of the product In a data science project, it is essential to build a maintainable pipeline because: The characteristics of data can change frequently Data…Data Science8 min readData Science8 min read
Published in Towards Data Science·Sep 22, 2022Member-only4 Tools to Automatically Extract Data from Datetime in PythonHow to Extract Datetime From Text and Data From Datetime — Motivation Datetime features are important for time series and forecasting. However, you don’t always have clean datetime features to work with. Wouldn’t it be nice if you can automatically extract datetime from text and data from datetime? In this article, you will learn 4 tools to do exactly that. datefinder: Automatically Find Dates and Time in a Python StringPython4 min readPython4 min read
