Research map / Learning theory and algorithms / Prediction and decision losses
DppNet: Approximating Determinantal Point Processes with Deep Networks
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 |
|---|---|---|---|
| Replace direct DPP sampling with DppNet sampling | DPP sampling on the unit square · Unit square, k=20 | NLL: 154.95 ± 2.93 -> 153.44 ± 2.07 | Helped |
| Use greedy mode sampling instead of sampling from the DPP | DppNet · MNIST, all classes, k=20 | NLL: 49.2 ± 0.1 -> 48.6 ± 0.2 | Helped |
| Use DppNet greedy mode sampling instead of direct DPP sampling | DPP sampling · CelebA, k=20 | NLL: 49.04 ± 2.03 -> 49.28 ± 1.57 | No clear effect |
| Use DppNet greedy mode sampling instead of direct DPP sampling | DPP sampling · MovieLens, k=20 | NLL: 84.29 ± 0.20 -> 80.21 ± 0.33 | Helped |
| Replace direct DPP sampling with DppNet sampling | DPP sampling · Ground set size 100, subset size 20, batch size 32 | Sampling time (seconds): 2.74 ± 0.02 seconds -> 0.10 ± 0.001 seconds | Helped |
| Remove the feed-forward network after inhibitive attention (ablation) | DppNet · MNIST, all classes, k=20 | NLL: 48.6 ± 0.2 -> 51.3 ± 0.1 | Hurt |
| Remove inhibitive attention and use a feed-forward network (ablation) | DppNet · MNIST, all classes, k=20 | NLL: 48.6 ± 0.2 -> 51.4 ± 0.1 | Hurt |
Closest papers by meaning
- Deep Learning of Determinantal Point Processes via Proper Spectral Sub-gradient (ICLR 2020 · similarity 0.91)
- State-of-art minibatches via novel DPP kernels: discretization, wavelets, and rough objectives (NeurIPS 2026 · similarity 0.91)
- Sampling from a k-DPP without looking at all items (NeurIPS 2020 · similarity 0.90)
- Kronecker Determinantal Point Processes (NeurIPS 2016 · similarity 0.89)
- Learning Determinantal Point Processes with Moments and Cycles (ICML 2017 · similarity 0.89)
- Small coresets via negative dependence: DPPs, linear statistics, and concentration (NeurIPS 2024 · similarity 0.87)
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.