Published
SVM Router
Support Vector Machine based routing — leverages learned decision boundaries for high-dimensional query spaces.
Overview
The SVM Router uses a Support Vector Machine classifier to route queries based on learned decision boundaries in high-dimensional embedding spaces.
How It Works
By mapping query embeddings into a high-dimensional feature space, the SVM finds the optimal hyperplane that separates queries destined for different LLMs. New queries are classified based on which side of the hyperplane they fall on.
Strategy
Uses SVM classifier to route queries based on learned decision boundaries.
API Endpoint
autoroute:svmrouter
Use Cases
- High-dimensional query spaces
- When clear decision boundaries exist between query types
- Scenarios with well-separated query categories
Best Practices
Kernel Selection
The choice of kernel function significantly impacts performance. RBF kernels work well for most text routing tasks, but linear kernels can be faster when the data is already well-separated in the original space.
Related Models
- KNN Router — Simpler alternative for lower-dimensional spaces
- MLP Router — For non-linear routing patterns