Pandas Pow Series

Here's the exponentiation node with actual personality. Pandas Pow Series raises a DataFrame to the power of a Series - meaning instead of one exponent for the whole table, you get a different exponent per column (or per row, if you arrange it that way). That's the middle ground between the boring scalar versions and the full DataFrame-to-DataFrame node: you're applying a vector of exponents to a table. Squash column A with 0.5, square column B with 2, leave column C alone with 1 - all in one step.

It's from HowToSD/ComfyUI-Data-Analysis, the pandas-as-nodes extension. This one needs a bit more setup than its siblings, because it expects a PDSERIES on one of its inputs - you'll typically build that series with Pandas Select Column As Series, Pandas Create, or one of the series-creating nodes. That makes it a two-node operation, which is exactly why people usually default to the scalar version. Fine - this one exists for when you genuinely need per-column powers.

How it works

Under the hood it's dataframe.pow(series). The interesting part is alignment: pandas matches the Series' labels against the DataFrame's column labels by default. So the Series' index labels should be your column names. Alignment means the two don't have to be in the same order, but anything that doesn't match produces NaN - which is the classic way this node goes wrong. If your output is a grid of missing values, your Series labels and column names don't line up.

Inputs and output

• a_dataframe - the base table.
• b_series - the PDSERIES of exponents, aligned to the DataFrame's columns.

Output is one DATAFRAME, same shape as the input wherever labels matched.

Installation

The whole pack installs identically. ComfyUI Manager: search Data analysis, install ComfyUI-Data-Analysis. Or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/HowToSD/ComfyUI-Data-Analysis.git
mv ComfyUI-Data-Analysis data-analysis   # examples need this folder name
pip install -r requirements.txt

Restart, reload. Deps are pandas, matplotlib, seaborn, scipy, scikit-learn, openpyxl, lxml - no GPU, no models. PyTorch stuff is a separate extension since March 2025.

The real gotcha

Label alignment is everything here. The Series' index must contain the column names you want to exponentiate, or you get NaN silence rather than an error. Also remember the transpose alternative: pandas can align a Series to rows instead if you arrange it that way. For most people, though, the honest advice is - if you find yourself fighting this node's alignment, ask whether the float-scalar version plus a Select Columns is actually all you needed. The Series version shines specifically when different columns genuinely need different exponents.