This Python library, tnlearn, enables the construction of neural networks using diverse, task-specific neurons. It targets researchers and practitioners seeking to enhance feature representation and model performance by moving beyond homogeneous neuron designs, inspired by the human brain's neuronal diversity. The core benefit is improved representational capacity through neurons intrinsically biased for specific tasks.

How It Works

Tnlearn employs vectorized symbolic regression to discover optimal mathematical formulas that fit input data. These formulas are then parameterized to create learnable, task-based neurons. This approach allows for neurons with intrinsic inductive biases tailored to specific tasks, enhancing feature representation within a given network architecture.

Quick Start & Requirements

Installation is straightforward via pip: pip install tnlearn. For GPU acceleration, ensure PyTorch >= 2.1.0 is installed, compatible with CUDA >= 12.1. Other major dependencies are installed automatically. Local installation can be done using pip install -e .. API documentation is available on Read the Docs.

Highlighted Details

• Features vectorized symbolic regression and polynomial tensor regression for generating task-based neurons.
• Includes DrSR, an LLM-based symbolic regression method for discovering interpretable expressions as task-based neurons.
• Benchmarks demonstrate superior performance (lower MSE) compared to XGBoost, LightGBM, CatBoost, TabNet, TabTransformer, FT-Transformer, and DANETs on particle collision and asteroid prediction datasets, with the Task-based Network achieving the lowest MSE.

Maintenance & Community

The project is a work by Meng Wang, Juntong Fan, Hanyu Pei, Tieyun LI, Jingxiao Liao, and Fenglei Fan. No specific community channels (e.g., Discord, Slack) or roadmaps are detailed in the README.

Licensing & Compatibility

Tnlearn is released under the Apache License 2.0, which is generally permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

The README does not explicitly detail limitations, alpha status, or known bugs. The DrSR feature requires specific LLM configurations and API keys. The library appears primarily focused on tabular data for regression tasks.