Research map / Optimization / Optimizers
Near-Universal Multiplicative Updates for Nonnegative Einsum Factorization
John Hood, Aaron Schein
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 |
|---|---|---|---|
| use NNEinFact multiplicative updates instead of Adam automatic differentiation | custom nonnegative tensor models · Uber, ICEWS, WITS heldout | heldout loss and time to convergence: less than half the test loss, up to 90 times faster | Improved |
| use custom tailored decompositions instead of CP/Tucker/tensor-train | NNEinFact · Uber, ICEWS, WITS heldout | heldout (alpha,beta)-divergence: over 37% better prediction | Improved |
| custom model vs CP at ICEWS alpha=0.8 | NNEinFact custom · ICEWS, alpha=0.8 | heldout loss: CP offers a 1% reduction | Hurt |
| multiplicative updates instead of HALS for CP and Tucker | CP, Tucker · Uber, ICEWS, WITS | heldout loss | Improved |
| low-rank Tucker instead of full Tucker (ablation) | Tucker · ICEWS, WITS | heldout loss | Mixed |
| NNEinFact vs Adam on two of six (alpha,beta) settings | NNEinFact · Uber | runtime and heldout loss: better on four of six settings | Mixed |
Closest papers by meaning
- SWIFT: Scalable Wasserstein Factorization for Sparse Nonnegative Tensors (AAAI 2021 · similarity 0.87)
- Provable Online CP/PARAFAC Decomposition of a Structured Tensor via Dictionary Learning (NeurIPS 2020 · similarity 0.85)
- Distributed Flexible Nonlinear Tensor Factorization (NeurIPS 2016 · similarity 0.85)
- Scaling and Scalability: Provable Nonconvex Low-Rank Tensor Estimation from Incomplete Measurements (JMLR 2022 · similarity 0.85)
- Implicit Regularization for Tubal Tensor Factorizations via Gradient Descent (ICML 2025 · similarity 0.84)
- Einsum Benchmark: Enabling the Development of Next-Generation Tensor Execution Engines (NeurIPS 2024 · similarity 0.84)
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.
