Practical machine learning
See the bigger picture in your grid search.
Go beyond the best score. Use dendrogram heatmaps to understand the shape of your hyperparameter search.
Notes from an ML practitioner
Machine learning.
Beyond the
notebook.
I'm Cristian Lungu. I explore how machine learning works — and help teams put it to work.
Practical experiments. Clear explanations. A healthy skepticism of the hype.
THE LEARNING PROCESSFIG. 01
At the intersection ofApplied machine learningEngineering & experimentationTeaching & consulting
A few good places to start
Ideas, with the working shown.
From first principles
What is cross entropy, really?
Build a loss function from scratch, compare implementations, and find out where the intuition breaks.
Interpretable models
A forest, explained by a single tree.
Explore model distillation: can a simpler decision tree capture what a random forest has learned?
From the notebook
Latest writing
3 min read
Finding substrings in a numpy vectorized function
Find a needle in a NumPy haystack: unpacking a vectorized approach to matching subsequences in large arrays.
3 min read
Connecting Colab to a local machine
Use the Colab notebook interface with your own compute. A practical guide to connecting a local runtime.
2 min read
Visualising grid search results
Look beyond the winning score. Explore hyperparameter relationships and model robustness with dendrogram heatmaps.
1 min read
A timer context manager in python
A small Python utility for measuring execution time, built around the context manager protocol.
The person behind the notebook
A practitioner. A teacher. Always a learner.
I work at the intersection of machine learning and software engineering. This is where I share the experiments, explanations, and lessons along the way.
A little more about meFrom reading to building
Have a machine learning
problem worth solving?
Let's turn the right questions into a practical way forward.