ML knowledge graph video

A machine-learning map that treats the field like a subway diagram — datasets, models, and tasks linked as nodes and edges — is getting fresh attention after Tivadar Danka posted a video version on April 17. (substack.com) Machine learning is a way to train software from examples instead of writing every rule by hand, and a knowledge graph is a map of concepts connected by labeled relationships. Danka’s March 30 post, “Explore Machine Learning as a Knowledge Graph,” framed the subject in exactly that format. (thepalindrome.org) The project sits inside Danka’s broader teaching catalog at The Palindrome, a Substack publication that describes itself as “mathematics ∪ machine learning,” and on his personal site, where he says he is a pure mathematician turned machine-learning engineer. His site also says he earned a PhD in 2016 and now focuses on “democratiz[ing] machine learning and mathematics.” (thepalindrome.org) (tivadardanka.com) The format answers a common beginner problem: machine learning is usually taught as separate lists of models, benchmarks, and math topics, even though practitioners move between them as connected parts of one workflow. Danka has made that same “connected” framing a recurring theme in other posts, including “Your Machine Learning Library” on August 10, 2025, and “The Single Most Undervalued Fact of Linear Algebra” on January 4, 2026. (thepalindrome.org 1) (thepalindrome.org 2) That matters in 2026 because the machine-learning toolkit is sprawling: supervised learning, reinforcement learning, neural networks, embeddings, evaluation metrics, and data pipelines often get introduced in different places and at different depths. A graph view compresses those pieces into one picture, showing which ideas sit upstream of others and which tasks share the same building blocks. (packtpub.com) (thepalindrome.org) Danka’s audience is already built around that kind of compression. Packt’s description of his 2025 book, *Mathematics of Machine Learning*, says he has “100k+ followers,” while The Palindrome says the newsletter reaches “tens of thousands of subscribers.” (packtpub.com) (thepalindrome.org) He has also been extending the graph idea beyond a single post. A recent Substack note said he was “finishing up with knowledge graphs and building more interactive tools,” and a separate Substack profile entry pointed to a plan for “an interactive knowledge graph explorer.” (thepalindrome.org) (substack.com) The video does not change machine-learning theory, but it packages the subject in a form that is easier to scan, revisit, and share. For learners staring at a crowded field, the point is the map itself. (substack.com) (thepalindrome.org)