Turns a handful of experimental runs into a statistically faithful dataset. The engine learns the joint distribution of your real experiments — not just per-column shapes, but the non-linear correlations between conditions and outcomes.
Choosing a mode
Matflow Neural — conditional-GAN-style mode for mixed continuous/categorical columns and multi-modal distributions. Strongest default for larger, more complex datasets.
Matflow Hybrid — GAN-style network combined with a copula transform; often more stable on small samples.
Matflow Statistical — fits marginals and a correlation structure directly. Fastest and the most defensible when rows number in the tens.
Inputs and settings
Dataset — your uploaded or demo dataset.
Rows to generate — how many synthetic rows you want (within tier limits).
Column types — verified automatically; override if the detector mislabeled a column.
Seed — set a fixed seed for reproducible runs.
Range clipping
Generated rows are clipped to the physically observed range of each column — a generator can otherwise extrapolate a percentage column past 0–100%. The clipping is applied after generation; constraints enforced natively are on the roadmap.
After generation
A synthetic dataset appears in your Datasets list, marked with source "synthetic". Run Evaluation on it before using it for modeling — that is the step that decides whether the augmentation is trustworthy.