equivariant-architecture-designer

Design neural network architectures respecting symmetry groups like E(3) and SO(3).

142|20|Updated Oct 22, 2025

Search Tags:#symmetry #geometric deep learning #equivariant neural networks #deep learning architecture #e3nn #escnn

Ready to run

npx skills add https://github.com/lyndonkl/claude --skill equivariant-architecture-designer

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: equivariant-architecture-designer
Source: https://github.com/lyndonkl/claude/tree/main/skills/equivariant-architecture-designer
Command: npx skills add https://github.com/lyndonkl/claude --skill equivariant-architecture-designer

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you design neural network architectures that inherently respect the symmetries of your data, leading to more efficient, robust, and generalizable models.

Core Features & Use Cases

Symmetry-Aware Architecture Design: Recommends network structures tailored to specific symmetry groups (e.g., rotations, translations, permutations).
Library and Layer Guidance: Suggests appropriate libraries (like e3nn, escnn) and equivariant layer types (convolutions, nonlinearities, normalization).
Use Case: When building a model for 3D molecular data where rotational and translational symmetry is crucial, this Skill will guide you in selecting E(3)-equivariant layers and designing a network topology that leverages these symmetries for better performance.

Quick Start

Use the equivariant-architecture-designer skill to design a neural network architecture for E(3) symmetry on point cloud data.

Frequently Asked Questions about equivariant-architecture-designer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I design a neural network that respects 3D rotational and translational symmetry?▼

To design a neural network respecting 3D symmetry, use E(3)-equivariant layers. This approach provides architecture patterns and specific libraries like e3nn to build models that inherently respect rotational and translational symmetries for improved efficiency.

What is the best way to choose equivariant nonlinearities and normalization layers for my architecture?▼

Choosing equivariant nonlinearities and normalization layers requires selecting components that preserve your target symmetry group. This Skill guides the selection of validated nonlinearities and normalization techniques tailored to groups like E(3), SO(3), C_n, and S_n.

When do I need equivariant neural networks for deep learning architecture design?▼

You need equivariant neural networks when your data has inherent symmetries like rotations, translations, or permutations. Building models that respect these validated symmetry groups improves model efficiency, robustness, and generalization on complex geometric data.

Does this equivariant architecture designer support libraries like e3nn and escnn?▼

Yes, this equivariant architecture designer supports libraries like e3nn and escnn. It provides specific library guidance and suggests appropriate equivariant layer types, including convolutions and nonlinearities, to implement your geometric deep learning model.

How can I build a model for 3D molecular data that requires rotational symmetry?▼

To build a model for 3D molecular data requiring rotational symmetry, select E(3)-equivariant layers. This Skill guides you in designing a network topology that leverages these symmetries, leading to better performance on point cloud data.

What are the limitations of using standard neural networks instead of equivariant architectures for geometric data?▼

Standard neural networks lack built-in symmetry awareness, reducing model efficiency and generalization on geometric data. Using equivariant architecture design ensures your model inherently respects validated symmetry groups like SO(3) and S_n, preventing unnecessary learned redundancies.