Adding to the MLFlow Pipeline - Conformal Prediction

As I'm building out my MYFlow pipeline, I'm looking for ML Models that are both interesting and valuable. Conformal Prediction is a hot topic, and I enjoy this informal introduction.

• https://www.youtube.com/watch?v=nql000Lu_iE

Their code:

• https://github.com/aangelopoulos/conformal-prediction

Their paper:

• https://arxiv.org/pdf/2107.07511

I haven't made formal New Year resolutions, but one is to spend at least one day a week reading papers in probability and machine learning.

Notes

Conformal predictions measure uncertainty in machine learning models. They provide a helpful way for the user to see the probability of whether a guess matches the real-world truth. It allows us to transform data from complex, high-dimensional spaces into spaces we can intuitively understand and analyze.

In my daily work, I see conformal mapping as a valuable tool to reveal clusters, patterns, and outliers for pricing analysis. By quantifying uncertainty, we can measure the appropriateness of a price given volatile conditions to guide competitive pricing that earns more money for each listing.