User guide

Short, ready-to-copy snippets, one for each task. Find your task below, or open a section in the menu.

Find a snippet by task

Start

• install the client → Install the client and check its version

• create an API key / log in → Create an API key, Log in to the server

• run a first avatarization → Launch my first avatarization

• set the seed, output format or PIA context of a run → Set the options of a run

Prepare the data

• upload a DataFrame or a file → Add a table from a DataFrame or a file

• force the type of a column → Change the type of a column

• remove a column → Drop a column

• handle a large dataset → Handle a large dataset

• anonymize linked tables → Anonymize several linked tables (multi-table)

• anonymize time series → Anonymize time series

Tune the avatarization

• start from the recommended parameters → Get the recommended parameters

• know when to use set_parameters or update_parameters → set_parameters or update_parameters?

• change k or ncp → Set k (privacy / utility trade-off), Set ncp (number of components)

• preserve a variable of interest better → Give more weight to a variable

• handle categories with many modalities → Handle categorical variables with many modalities

• keep a column out of the modelling (freeze, similarity) → How to keep a column out of anonymization

• handle missing values → Choose the imputation of missing values, Get avatars without missing values

• reproduce a run exactly → Reproduce a run exactly

Augment, add DP, pseudonymize

• generate more rows / rebalance a class → Augment your dataset, Rebalance a minority class

• use differential privacy → Use differential privacy (FastDP), Use differential privacy (OpenDP, MST)

• replace names, emails, IDs → Pseudonymize PII columns (Beta)

Keep business rules

• keep a total equal to the sum of its parts, or A ≤ B → Keep a total equal to the sum of its parts

• keep start and end dates coherent → Keep start and end dates coherent across records

• group rare modalities → How to group rare modalities

• keep a discrete variable discrete → Keep a discrete numeric variable discrete

• write my own processor → Write my own processor

Evaluate

• define what an attacker knows → Define the attack scenario

• read the metrics → Read the metrics

• evaluate a synthetic dataset made elsewhere → How to evaluate an existing synthetic dataset

Run and get results

• run only some jobs → Run all or some of the jobs

• follow, debug or cancel a job → Follow a job, Read the error of a failed job, Cancel jobs

• get the avatars → Get the avatars

• show plots / download the reports → Render plots, Download the reports

Reuse

• reload a past run (also from the web app) → Reload results from a previous run

• export or replay a YAML configuration → Export a configuration as YAML, Launch a job from a YAML file

• continue in Python a job launched in the web app → Continue in Python a job launched in the web app

Something does not work? See Troubleshooting.