nschaetti/EchoTorch

A Python toolkit for Reservoir Computing and Echo State Network experimentation based on pyTorch. EchoTorch is the only Python module available to easily create Deep Reservoir Computing models.

What it solves

EchoTorch provides a research-oriented framework for implementing and testing various flavors of Echo State Networks (ESNs) and Reservoir Computing. It simplifies the creation of these models, which are typically used for time-series prediction and pattern recognition, while allowing them to be integrated into larger deep learning architectures.

How it works

Built on PyTorch, the library organizes its functionality into several modules:

  • Models and NN: Provides ready-to-train ESN components and neural network layers specifically for reservoir computing.
  • Utilities: Includes tools for matrix generation (Gaussian, uniform, etc.), hyperparameter optimization (genetic, grid, and random search), and visualization.
  • Conceptors: Implements neural filters (conceptors) for pattern generation, morphing, and memory management.
  • Evaluation: Offers tools for cross-validation and statistical testing to compare model performance.
  • Datasets: Includes pre-built datasets for common ESN tasks like NARMA and MNIST.

Who it’s for

Researchers and developers interested in reservoir computing and Echo State Networks who want a PyTorch-compatible implementation for experimentation and academic study.

Highlights

  • PyTorch Integration: Designed to be integrated into deep architectures for future research.
  • Conceptor Support: Advanced capabilities for learning, generating, and morphing patterns using conceptor-based neural filtering.
  • Comprehensive Tooling: Built-in support for hyperparameter optimization and data transformation specific to ESNs.
  • Diverse Examples: Extensive examples covering time-series prediction (Mackey-Glass, NARMA), image classification (MNIST), and unsupervised learning.

Related

  • Project
    reservoirpy/reservoirpy

    A flexible Python library for implementing Reservoir Computing architectures like Echo State Networks for time-series prediction and classification.

  • Project
    neuralhydrology/neuralhydrology

    A PyTorch-based Python library designed for deep learning research in hydrology, allowing users to train neural networks via configuration files.

  • Project
    pytorch/pytorch

    A Python-based machine learning framework that provides GPU-accelerated tensor computation and a dynamic autograd system for building deep neural networks.

  • Project
    sktime/pytorch-forecasting

    A PyTorch-based library for time series forecasting that provides a high-level API and state-of-the-art deep learning architectures like Temporal Fusion Transformers and N-BEATS.

  • Project
    facebookresearch/pytorch3d

    A PyTorch-based library for 3D Computer Vision research providing differentiable rendering and efficient tools for manipulating 3D meshes and point clouds.