Defining Parameter Data

KKT-HardNet currently uses CSV datasets for parameter samples and, when needed, supervised variable targets.

CSV-based Dataset

Attach data with:

model.dataset(parameters="parameters.csv", variables="variables.csv")

The parameters.csv columns must match the names passed to add_parameter(...). The variables.csv columns must match the names passed to add_variable(...).

Data Requirements by Workflow

Workflow	Required CSV files	Description
`model()`	`parameters.csv` and `variables.csv`	Supervised surrogate learning from known solutions.
`estimate()`	`parameters.csv` and `variables.csv`	Inverse estimation using observed variables.
`optimize()`	`parameters.csv`	Unsupervised optimization over supplied parameter samples.

Example

parameters.csv

x1,x2
0.0,0.0
0.5,-0.25
-0.8,0.4

variables.csv

y1,y2,y3
0.2,0.1,-0.1
0.4,0.0,-0.3
-0.1,0.5,0.2

Then attach the files:

model.dataset(
    parameters="parameters.csv",
    variables="variables.csv",
)

Notes

CSV headers are required.
Header names should be unique.
The row counts of parameters.csv and variables.csv must match when both files are provided.
The model does not automatically reject infeasible parameter samples before training.