The ISIDA/GTM software package is a comprehensive toolkit designed for the generation, analysis, and visualization of Generative Topographic Maps (GTM). This package implements the GTM algorithm, as originally introduced by Bishop et al., along with its incremental, multinomial noise model, and block-based training [4] variations. It is intended for researchers, data scientists, and practitioners who require advanced tools for mapping, classifying, and regressing high-dimensional data, particularly in the field of cheminformatics.
- GTMapTool: The primary software for building GTM models, continuing training, and projecting new data onto existing maps.
- GTMClass: A tool for classification tasks, enabling the generation of property landscapes for labeled datasets and defining applicability domains (AD).
- GTMReg: A tool for regression tasks, enabling the generation of property landscapes for labeled datasets and defining applicability domains (AD).
- GTMReSample: Allows resampling of GTM models to adjust node resolution without the need for retraining.
The ISIDA/GTM package is designed to:
- Visualize high-dimensional data in 2D or 3D latent spaces.
- Perform Classification and Regresssion of properties of chemical compounds or other complex datasets.
- Define applicability domains to ensure reliable predictions.
- Facilitate integration with workflows such as KNIME or Pipeline Pilot for streamlined data analysis.
For compilation, the ISIDA/Kernel project is required.
For detailed usage instructions, command-line options, and examples, please refer to the ISIDA/GTM Manual.