Research map / Probabilistic and scientific modeling / Molecules and proteins
User-Dependent Neural Sequence Models for Continuous-Time Event Data
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 user-specific latent embeddings inferred with a variational mixture of experts | NHP and RMTPP · Meme, Reddit, Amazon, and LastFM | mean predicted mark rank: Meme: 283 -> 225 (NHP); 159 -> 143 (RMTPP). Reddit: 35.2 -> 19.4 (NHP); 35.6 -> 20.0 (RMTPP). Amazon: 22.8 -> 21.6 (NHP) | Helped |
| Condition on inferred user-specific latent embeddings | NHP and RMTPP · Meme, Reddit, Amazon, and LastFM | source identification error rate: Meme: 9.18% -> 5.04% (MoE-NHP), 4.64% (MoE-RMTPP); Reddit: 3.24% -> 1.38%, 2.02%; Amazon: 12.88% -> 8.34%, 9.94%; LastFM | Helped |
| Add user-specific latent embeddings | NHP and RMTPP · Meme | mean L1 loss for predicted time: 15.93 -> 15.64 (NHP); 14.01 -> 12.97 (RMTPP) | Helped |
| Add user-specific latent embeddings | NHP and RMTPP · Reddit | mean L1 loss for predicted time: 4.51 -> 4.57 (NHP); 3.87 -> 3.88 (RMTPP) | Mixed |
| Add user-specific latent embeddings | NHP and RMTPP · Amazon | mean L1 loss for predicted time: 2.43 -> 2.39 (NHP); 2.06 -> 1.97 (RMTPP) | Helped |
| Add user-specific latent embeddings | NHP and RMTPP · LastFM | mean L1 loss for predicted time: 0.61 -> 0.56 (NHP); 0.32 -> 0.33 (RMTPP) | Mixed |
| Use personalized mixture-of-experts models instead of decoder-only models | NHP and RMTPP · all four datasets | held-out log-likelihood | Helped |
| Use personalized mixture-of-experts models instead of decoder-only models | NHP and RMTPP · early sequence predictions | mark cross entropy | Helped |
| Personalize event timing predictions | NHP and RMTPP · all four datasets | predicted-time performance | No clear effect |
Closest papers by meaning
- EasyTPP: Towards Open Benchmarking Temporal Point Processes (ICLR 2024 · similarity 0.88)
- A Variational Autoencoder for Neural Temporal Point Processes with Dynamic Latent Graphs (AAAI 2024 · similarity 0.88)
- A Variational Point Process Model for Social Event Sequences (AAAI 2020 · similarity 0.87)
- Latent LSTM Allocation: Joint Clustering and Non-Linear Dynamic Modeling of Sequence Data (ICML 2017 · similarity 0.87)
- Interacting Diffusion Processes for Event Sequence Forecasting (ICML 2024 · similarity 0.87)
- Deep Continuous-Time State-Space Models for Marked Event Sequences (NeurIPS 2025 · similarity 0.86)
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.