Data Scientists, Don’t Overlook log1p

Author(s): Suraj Jha

Originally published on Towards AI.

Master Smoother Transformations and Tame Exponential Data Like a Pro

This member-only story is on us. Upgrade to access all of Medium.

Data Scaling is one of the crucial steps for preparing dataset for machine learning model.

You might have work with techniques like normalization, standardization, logarithmic transformations, as they are well recognized in data science field.

But have you ever used log1p? or did you ever hear it before, probably not most of us don’t hear anything about it.

Why log1p is so special yet underrated?

What are its practical applications and why every data scientist should know about this?

Lets get our answers, shall we.

log1p() is a function that calculates the natural logarithm of (1+x)

At first you might confuse it with another logarithmic useless function but it's not.

The ability to handle zeros, small values and negative values so gracefully that making it superior from those traditional logarithms.

Remember when you use logarithmic transformations to reduce the skewness in data. However, the logarithm of 0 is still undefined, so any value closer to zero gets us the output as negative.

That’s when log1p() comes into the picture, it fixes this issue by shifting the input (adding 1). It ensures the stability even if we have zero value or small value.

Avoids Negative Infinity

Suppose you have dataset… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI