Research map / Label-efficient and robust learning / Recognition and long-tail
Enhancing Binary Encoded Crime Linkage Analysis Using Siamese Network
Yicheng Zhan, Fahim Ahmed, Amy Burrell, Matthew Tonkin, Sarah Galambos, Jessica Woodhams, Dalal Alrajeh
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| integrate geo&temp at decoder instead of input concat (ablation) | MLP Siamese Autoencoder · ViCLAS multiple (no data reduction), architecture study | AUC (%): 76.43 -> 77.29 | Improved |
| integrate geo&temp at decoder instead of input concat (ablation) | 1D CNN Siamese · ViCLAS multiple (no data reduction), architecture study | AUC (%): 58.45 -> 61.74 | Improved |
| integrate geo&temp at decoder instead of input concat (ablation) | SIREN Siamese · ViCLAS multiple (no data reduction), architecture study | AUC (%): 55.19 -> 58.28 | Improved |
| add skip connections (ablation) | Siamese Autoencoder · ViCLAS multiple (no data reduction), architecture study | AUC (%): 63.03 -> 56.48 | Hurt |
| use MLP encoder instead of 1D CNN or SIREN (ablation) | Siamese Autoencoder · ViCLAS multiple (no data reduction), architecture study | AUC (%): 57.10 -> 65.30 | Improved |
| use 2+2 layers instead of 1+1 (ablation) | MLP Siamese Autoencoder · ViCLAS multiple (no data reduction), architecture study | AUC (%): 52.49 -> 77.29 | Improved |
| use 2+2 layers instead of 4+4 (ablation) | MLP Siamese Autoencoder · ViCLAS multiple (no data reduction), architecture study | AUC (%): 63.57 -> 77.29 | Improved |
| add map 1 data reduction (ablation) | Siamese Autoencoder · ViCLAS multiple victim-offender-scene | AUC (%): 77 -> 83 | Improved |
| add map 5 data reduction (ablation) | Siamese Autoencoder · ViCLAS multiple victim-offender-scene | AUC (%): 77 -> 84 | Improved |
| add map 2 data reduction (ablation) | Siamese Autoencoder · ViCLAS multiple victim-offender-scene | AUC (%): 77 -> 77 | No clear effect |
| add map 4 data reduction (ablation) | Siamese Autoencoder · ViCLAS multiple victim-offender-scene | AUC (%): 77 -> 76 | No clear effect |
| Siamese Autoencoder with decoder geo&temp instead of Naive Siamese | Naive Siamese · ViCLAS multiple victim-offender-scene | AUC (%): 76 -> 77 | Improved |
Closest papers by meaning
- Spatial-Temporal Augmentation for Crime Prediction (Student Abstract) (AAAI 2024 · similarity 0.82)
- Hawkes-Enhanced Spatial-Temporal Hypergraph Contrastive Learning Based on Criminal Correlations (AAAI 2024 · similarity 0.81)
- T-NET: Weakly Supervised Graph Learning for Combatting Human Trafficking (AAAI 2024 · similarity 0.80)
- Barely Supervised Learning for Graph-Based Fraud Detection (AAAI 2024 · similarity 0.78)
- Seeing What Matters: Generalizable AI-generated Video Detection with Forensic-Oriented Augmentation (NeurIPS 2025 · similarity 0.77)
- Community Forensics: Using Thousands of Generators to Train Fake Image Detectors (CVPR 2025 · similarity 0.77)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
