Tutorials#

Comprehensive step-by-step guides that walk through complete workflows and real-world scenarios. These tutorials help you learn TanaT systematically by combining multiple concepts into practical examples.

Building & Ingesting Data#

Master the builder API and load data from heterogeneous sources.

Building pools from multiple sources: combine Parquet, CSV, and SQL sources into a single pool, configure builder options, compose trajectory pools, and manage the workspace.

MIMIC-IV: Clinical Cohort Analysis#

Real patient cohort workflows from the MIMIC-IV demo dataset.

Exploring a patient cohort: load, inspect, visualise, train/test split.
Filtering and preparing a cohort: criteria, T0 anchor, relative window.
Analysing and clustering a cohort: distance matrix, hierarchical clustering, faceted timeline.
Survival analysis by admission cluster: mortality endpoint, survival target, per-cluster Kaplan-Meier curves.

Education: Learning Session Analysis#

Sequence analysis workflows for MOOC session data.

Exploring learner activity sequences: load, inspect, visualise, train/test split.
Clustering sessions by action patterns: session clustering with criteria and temporal patterns.

Time Series to Sequences#

Convert raw time series into TanaT sequence pools with Aeon.

Discretizing time series into sequences: apply segmentation and quantization methods to ingest raw time series as TanaT sequences.

Deep Learning with TanaT#

End-to-end deep learning workflows using TanaT with PyTorch, SWoTTeD, and Fed-BioMed.

Learning clinical phenotypes with SWoTTeD: OHE tensor, dictionary learning, phenotype interpretation.
Federated learning with TanaT: use Fed-BioMed from data preparation with TanaT.