Differentiable Clustering

Differentiable clustering integrates clustering algorithms into differentiable frameworks, enabling their use within larger, end-to-end trainable neural networks. Current research focuses on developing efficient and scalable differentiable versions of classic algorithms like k-means, as well as exploring novel approaches based on spanning forests and associative memories. This allows for improved performance in various applications, including neural network compression, causal graph discovery, and robotics (e.g., lidar odometry), by enabling the optimization of clustering parameters alongside other model parameters. The resulting methods offer enhanced accuracy and efficiency compared to traditional, non-differentiable clustering techniques.

Papers

June 13, 2024

Scalable and Flexible Causal Discovery with an Efficient Test for Adjacency
Alan Nawzad Amin, Andrew Gordon Wilson
Causal Discovery Causal Graph Potential Scalability Adjacency Hopping Differentiable Clustering

December 12, 2023

IDKM: Memory Efficient Neural Network Quantization via Implicit, Differentiable k-Means
Sean Jaffe, Ambuj K. Singh, Francesco Bullo
K$ Mean Implicit Language Weight Quantization Neural Network Quantization Differentiable Clustering

June 5, 2023

End-to-end Differentiable Clustering with Associative Memories
Bishwajit Saha, Dmitry Krotov, Mohammed J. Zaki, Parikshit Ram
Associative Memory Differentiable Neural Associative Memory Model Clustering Loss Differentiable Clustering

May 25, 2023

Differentiable Clustering with Perturbed Spanning Forests
Lawrence Stewart, Francis S Bach, Felipe Llinares López, Quentin Berthet
Semi Supervised Clustering Random Perturbation Efficient Gradient Differentiable Clustering

September 20, 2022

WGICP: Differentiable Weighted GICP-Based Lidar Odometry
Sanghyun Son, Jing Liang, Ming Lin, Dinesh Manocha
Point Cloud Data Closed Form Differentiable Expression Iterative Closest Point Efficient Slam Differentiable Clustering

Differentiable Clustering

Papers

Scalable and Flexible Causal Discovery with an Efficient Test for Adjacency

IDKM: Memory Efficient Neural Network Quantization via Implicit, Differentiable k-Means

End-to-end Differentiable Clustering with Associative Memories

Differentiable Clustering with Perturbed Spanning Forests

WGICP: Differentiable Weighted GICP-Based Lidar Odometry