Research map / Probabilistic and scientific modeling / Molecules and proteins
A Multi-Channel Neural Graphical Event Model with Negative Evidence
Tian Gao, Dharmashankar Subramanian, Karthikeyan Shanmugam, Debarun Bhattacharjya, Nicholas Mattei
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 |
|---|---|---|---|
| Add fake epochs to reinforce intervals with no observed events | MCN-GEM · PGEM synthetic datasets | log-likelihood | Improved |
| Add fake epochs | MCN-GEM · ICEWS Argentina, Brazil, Colombia, Mexico | log-likelihood | Improved |
| Add fake epochs | MCN-GEM · ICEWS Venezuela | log-likelihood | Hurt |
| Add one fake epoch instead of none (ablation) | MCN-GEM · ICEWS Argentina | log-likelihood: -3559.5 -> -2677.9 | Improved |
| Use two fake epochs instead of one (ablation) | MCN-GEM · ICEWS Argentina | log-likelihood: -2677.9 -> -2652.3 | Improved |
| Use three, four, or five fake epochs instead of one (ablation) | MCN-GEM · ICEWS Argentina | log-likelihood: -2677.9 -> -2772.1; -2803.1; -2881.0 | Hurt |
| Use multi-channel rather than single-channel hidden states (ablation) | MCN-GEM · ICEWS Argentina | log-likelihood | Improved |
| Add one fake epoch to the single-channel version (ablation) | MCN-GEM single-channel variant · ICEWS Argentina | log-likelihood: -4323.8±35.4 -> -3839.4.9±25.4 | Improved |
| Add one fake epoch to the multi-channel version (ablation) | MCN-GEM · ICEWS Argentina | log-likelihood: -4282.2±58.1 -> -3302.6±28.1 | Improved |
| Add attention with memory size 1, 2, or 3 (ablation) | MCN-GEM with one fake epoch · ICEWS Argentina | log-likelihood: -3302.6±28.1 -> -2841.0±23.5; -2693.9± 27.0; -2682.6±29.3 | Improved |
| Increase attention memory size from 3 to 5 or 10 (ablation) | MCN-GEM with one fake epoch · ICEWS Argentina | log-likelihood: -2682.6±29.3 -> -2871.8±126.9; -2947.3 ± 190.0 | Hurt |
| Use the full MCN-GEM model instead of MCN-noF | MCN-GEM · MIMIC | log-likelihood: -1380.4 -> -837.4 | Improved |
Closest papers by meaning
- User-Dependent Neural Sequence Models for Continuous-Time Event Data (NeurIPS 2020 · similarity 0.85)
- Modeling the Intensity Function of Point Process Via Recurrent Neural Networks (AAAI 2017 · similarity 0.85)
- In-Context Learning of Temporal Point Processes with Foundation Inference Models (ICLR 2026 · similarity 0.84)
- A Variational Autoencoder for Neural Temporal Point Processes with Dynamic Latent Graphs (AAAI 2024 · similarity 0.83)
- Residual TPP: A Unified Lightweight Approach for Event Stream Data Analysis (ICML 2025 · similarity 0.82)
- Interacting Diffusion Processes for Event Sequence Forecasting (ICML 2024 · similarity 0.81)
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.
