CLUSTERING • IMMUNE-INSPIRED COMPUTING

TRCC: Tissue Response & Cytokine Clustering

A novel clustering algorithm inspired by immune system cytokine signaling, designed to handle non-convex clusters and noise better than K-Means or DBSCAN.

Algorithm Formulation

Each point emits a cytokine signal inversely proportional to distance, weighted by density:

$$ S_i = \sum_{j \in N_\epsilon(i)} \exp\Big(-\frac{d(x_i, x_j)^2}{2\sigma^2}\Big) \cdot w_j $$

where $ S_i $ is the cytokine signal at point $ i $, $ d(x_i, x_j) $ is the Euclidean distance, $ \sigma $ is the sensitivity factor, and $ w_j $ is a density weight.

Clusters are merged if centroid distance is below a threshold:

$$ d(C_a, C_b) < D_{\max} $$

Results

Synthetic Datasets: Correctly separated spirals, moons, and diamond clusters where K-Means failed.
Car-Hacking IDS: TRCC + feature engineering improved anomaly clustering by 12% purity over DBSCAN.
RNA Expression: Achieved superior silhouette scores on gene expression datasets.

Silhouette(TRCC) = 0.71 vs DBSCAN = 0.52

Impact

TRCC opens new pathways for immune-inspired ML, with applications in cybersecurity IDS, biomedical clustering, and anomaly detection.