💡Numberwang Model Decides if Numbers Are Numberwang
A tiny model that knows if your number is Numberwang
TL;DR
The Numberwang model, a 1.8 MB JSON file, decides if numbers are Numberwang. It handles 11 languages, Roman numerals, and even fictional numbers with 88.9% accuracy. Perfect for anyone who needs to know if their number is truly Numberwang.
The Numberwang model, a 1.8 MB JSON file with 100 lines of Python code, determines if a number is Numberwang. This tiny model can handle inputs in 11 languages, Roman numerals, and even fictional numbers with 88.9% accuracy. If you're working with numerical data and need to categorize numbers, this model could be a fun and quirky tool to add to your workflow. The model's verdicts are based on the input's properties, not the language, and it memorizes rather than computes, making it a unique addition to your data processing toolkit. The model's accuracy for 'not Numberwang' is 0.820, for 'Numberwang' is 0.919, for 'not a number' is 0.951, and for 'Wangernumb' is 0.968.
Key Points
The Numberwang model is a 1.8 MB JSON file with 100 lines of Python code.
Handles inputs in 11 languages, Roman numerals, and fictional numbers.
Accuracy is 88.9% over 486 held-out adjudications, with macro-F1 of 0.896.
Verdicts include 'That's not Numberwang', 'THAT'S NUMBERWANG!', and more.
Model's architecture includes chars → Embedding(32) → Conv1d(128, k3) → ReLU.
Why It Matters
If you're working with numerical data and need to categorize numbers, the Numberwang model could be a fun and quirky tool to add to your workflow. It handles 11 languages, Roman numerals, and even fictional numbers with 88.9% accuracy. The model's verdicts are based on the input's properties, not the language, making it a unique addition to your data processing toolkit.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,482 builders reading daily.