Pandas Mul Series

Sometimes you don't want to multiply by one number, and you don't have a second whole table either - you've got one row of weights, or one column of factors, and you want it applied across a DataFrame. Pandas Mul Series is the middle option: multiply a DataFrame by a single pandas Series, element by element, aligning on labels. Weight each column by its own factor, or scale every row by a matching vector - that's this node's lane.

It comes from ComfyUI-Data-Analysis by Hide Inada (HowToSD), the pack that wraps pandas, Seaborn, and Matplotlib into ComfyUI nodes. It's a thin wrapper over pandas' DataFrame.mul(series), sitting between the scalar multiply nodes and the frame-vs-frame Pandas Mul.

How it works

The node runs a_dataframe.mul(b_series). pandas aligns the Series against the DataFrame's labels - typically by column name - and multiplies each column by its matching Series value. Where labels don't match, you get NaN. So a Series indexed by column names gives you per-column scaling; a Series aligned by row index gives per-row scaling. Alignment is doing the real work here, and alignment is also where things go wrong.

The practical path: get a Series out of Pandas Mean, Pandas Max, or Pandas Loc Row Series, feed it here, and you've normalized or weighted a whole table by one row's worth of numbers. That's the trick people actually use this for - divide (or multiply) every column by its own mean without building a second table.

The inputs that matter

• a_dataframe - the frame to scale.
• b_series - a PDSERIES providing the per-label factors.

Output is a DATAFRAME, same shape, labels aligned.

How to install it

Same pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/HowToSD/ComfyUI-Data-Analysis.git
mv ComfyUI-Data-Analysis data-analysis
pip install -r requirements.txt

or ComfyUI Manager → search "Data analysis" → install → restart → reload. Standard pack dependencies (pandas, matplotlib, seaborn, scipy, scikit-learn); no GPU, no models.

Common issues

• NaN columns that should have values - the Series' index doesn't line up with the DataFrame's labels. Check that the Series is indexed by the same names as your columns (or rows), depending on what you're scaling.
• Wrong direction - a Series indexed by column names scales per-column; indexed by rows, per-row. Build the Series with the alignment you intend.
• Mixing up siblings - scalar for a constant, this node for a vector of factors, frame-vs-frame for two full tables. Wrong pick = wrong shape of result.

This is the node that makes "normalize each column by its own mean" a two-node pipeline instead of a detour through a script. Learn the alignment rule and it's a workhorse.