Forecasting with NHiTs: Uniting Deep Learning + Signal Processing Theory for Superior Accuracy | by Nikos Kafritsas | Oct, 2024

NHiTS was revealed 2 years in the past — and has gained important consideration within the forecasting neighborhood since then.

First, it’s a flexible mannequin — accepting previous observations, future identified inputs, and static exogenous variables. It may be utilized throughout varied forecasting domains, together with power demand, retail, and monetary markets.

It’s light-weight, but high-performance. Not like typical DL fashions that depend on “slapping“ hidden layers, this mannequin leverages sign idea ideas to spice up efficiency with minimal parameters.

Lastly, its multi-rate sign sampling technique permits the mannequin to seize complicated frequency patterns — important for areas like monetary forecasting. The mannequin can be utilized for probabilistic forecasting too.

On this article, we’ll clarify NHiTS intimately, analyze its structure, and spotlight its strengths and internal workings with sensible examples.

Let’s get began.

✅ Discover the hands-on venture for NHITS within the AI Projects folder (Venture 2) of my publication, together with different cool initiatives! Don’t overlook to subscribe!