Points of interest in visual context.

? N-dimensional dataset
Topology mapping
Extracted topology map
Historical patterns
Projection of points of interest

Technology & Mission

We use Self-Organizing Maps (SOM) to learn and observe the representation of high-dimensional datasets that are challenging to visualize otherwise — revealing data topology to put complex observations in visual context.

"Bianor builds upon Lilypond, the visualization tool we developed to create enhanced Self-Organizing Map (SOM) displays. Lilypond was inspired by water lilies; as you can see, the extracted maps look like clusters of lily pads floating on the water's surface. Following this metaphor, Bianor is more like a scenery of descending butterflies, where we can observe which historical pattern they land on and whether it seems out of context." — Mate Balogh, Lead at Bianor, Ophelia R&D

Applications

Discover how Bianor is being used to explore high-dimensional data.

Bianor Use Case - Output 1 Bianor Use Case - Output 2
Case Study #01

Demonstration of the Lilypond Library

To demonstrate the capabilities of the Lilypond library, experiments were conducted across manifold learning, supervised and unsupervised analysis, and anomaly contextualization. The resulting 2D representations were interactively inspected through a dedicated Bianor instance.

View Experiments Inspect in Bianor

Get Started

Recommended for individual researchers, data scientists, and curious minds seeking a lightweight, zero-cloud setup directly on their local machine.

  1. Clone the ophelia-rnd/bianor repository and run it
  2. Train an SomRepresentation model on the desired dataset and save it via MLflow
  3. Link your representation through the catalog.yaml file
  4. Use the Bianor application to interactively inspect your data's two-dimensional representation

Our Ecosystem: Foundation & Frontier

From exploratory data analysis to automated contextual insights.

Community Edition

High-dimensional observations mapped to visual contexts in 2D.

  • SomRepresentation models tracked in MLflow
  • Interactive visualizer with various layers and options
    powered by lilypond
  • Open-source
  • Databricks application (optional)

Designed for data scientists, researchers — and curious minds.

Clone the repository Documentation

Biome Edition

Advanced platform for hierarchical views and context extraction.

  • All Community features
  • Advanced interactive features
  • Hierarchical view of representations and projections
  • Instant and detailed contextualization

Designed for advanced workflows and custom requirements.

Contact the Team