Use case based notebooks.
On this page:
This notebook is designed to help users successfully train synthetic models on complex datasets with high row and column counts. The code works by intelligently dividing a dataset into a set of smaller datasets of correlated columns that can be parallelized and then joined together.
Walk through the basics of using Gretel's Python SDK to create a synthetic dataset from a Pandas DataFrame or CSV.
Train a synthetic model locally and generate data in your environment.
Conditional data generation (seeding a model) is helpful when you want to preserve some of the original row data (primary keys, dates, important categorical data) in synthetic datasets.
Balance demographic representation bias in a healthcare set using conditional data generation with a synthetic model.
Use a synthetic model to boost the representation of an extreme minority class in a dataset by incorporating features from nearest neighbors.
Run a sweep to automate hyper parameter optimization for a synthetic model using Weights and Biases.
Augment a popular machine learning dataset with synthetic data to improve downstream accuracy and algorithmic fairness.
Measure the effects of different differential privacy settings on a model's ability to memorize and replay secrets in a dataset.
This notebook shows how to generate synthetic data directly from a multi-table relational database to support data augmentation and subsetting use cases.
Generate realistic but synthetic text examples using an open-source implementation of the GPT-3 architecture.
Generate synthetic daily oil price data using the DoppelGANger GAN for time-series data.
Produce a quality score and detailed report for any synthetic dataset vs. real world data.
Use Gretel ACTGAN model to conditionally generate additional minority samples on a dataset that only has a few instances of the minority class
Synthesize a sample database using Gretel Relational Synthetics
In this blueprint, we will create a transform policy to identify and redact or replace PII with fake values. We will then use the SDK to transform a dataset and examine the results.
Label and transform sensitive data locally in your environment.
In this deep dive, we will walk through some of the more advanced features to de-identify data with the Transform API, including bucketing, date shifts, masking, and entity replacements.
This notebook walks through creating a policy using the Transform API to de-identify and anonymize data in a Postgres database for test use cases.
This notebook uses Gretel Relational Transform model to redact PII in a sample database.
In this blueprint, we will create a classification policy to identify PII as well as a custom regular expression. We will then use the SDK to classify data and examine the results.
Label managed and custom data types locally in your environment.
In this blueprint, we analyze and label a set of freetext email dumps looking for PII and other potentially sensitive information using NLP.
In this notebook, we benchmark datasets and models to analyze multiple synthetic generation algorithms (including, but not limited to, Gretel models). The Benchmark report provides Synthetic Data Quality Score (SQS) for each generated synthetic dataset, as well as train time, generate time, and total runtime (in secs).
Evaluate synthetic data vs. real data trained on AutoML classifiers. The Gretel Synthetic Data Utility Report provides a detailed table of classification metrics.
Evaluate synthetic data vs. real data trained on AutoML regression models. The Gretel Synthetic Data Utility Report provides a detailed table of regression metrics.