An interactive learning platform

A learning platform for synthetic data research.

OmicsHub Space is an interactive learning platform, through techniques used in current practice, to original work from our lab.

Start with the basics → Browse research

At a glance

Basic modules

Foundations every researcher needs first.

Advanced modules

Techniques at the level of current practice.

Research modules

Methods and results from our own lab.

The curriculum

Three tiers, one path.

Pick a tier that matches where you already are. Each module is self-contained, but the order on this page is the order we'd recommend to a new researcher joining the lab.

01 / 03

Basic

First principles. The vocabulary and intuitions you need before anything else makes sense.

B01 · Basic Open

Basics of Synthetic Data

What it is, why we make it, how we make it, and whether it's any good.

Random oversampling, SMOTE, and ADASYN — when to use each and what can go wrong.

How to read, analyze run oversampling models.

02 / 03

Advanced

Working methods used in modern papers and pipelines. Builds directly on the basics.

A01 · AdvancedOpen

Machine Learning

Supervised learning, evaluation, and the failure modes nobody puts in the introduction.

A02 · AdvancedComing soon

Under-sampling Methods

Random under sampling, edited nearest neighbor, Tomek-links.

lessons6

min50

levelintro

In preparation

theorycoding

A02 · AdvancedComing soon

Deep Generative Models

VAEs, GANs, normalising flows, and diffusion — chosen as if you had to defend the choice.

lessons9

min95

leveladvanced

In preparation

theorycoding

A03 · AdvancedComing soon

Tabular Data with CTGAN

Mode-specific normalisation, conditional sampling, and CTGAN end-to-end on real data.

lessons7

min75

leveladvanced

In preparation

coding

A04 · AdvancedComing soon

Differential Privacy

ε, δ, and the mathematics of admitting how much your data can leak.

lessons8

min80

leveladvanced

In preparation

theoryethics

A05 · AdvancedComing soon

Evaluation in Practice

TSTR, membership inference, fidelity scores — and how to read them without lying to yourself.

lessons7

min70

leveladvanced

In preparation

interactivecoding

03 / 03

Research result

Methods and findings from our own lab — packaged so others can learn from and build on them.

R01 · ResearchOpen

CTGAN-ENN

a hybrid data-level framework that combines conditional tabular GAN synthesis with Edited Nearest Neighbour cleaning to simultaneously address imbalance and overlap in customer churn datasets.