Low-Rank Factorization of Determinantal Point Processes
Authors: Mike Gartrell, Ulrich Paquet, Noam Koenigstein
AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We compare the performance of our low-rank DPP model to a full-rank DPP on two real-world datasets in Section 3. In this section we compare the low-rank DPP model with a full-rank DPP that uses a fixed-point optimization algorithm called Picard iteration (Mariet and Sra 2015) for learning. |
| Researcher Affiliation | Industry | Mike Gartrell Microsoft mike.gartrell@acm.org Ulrich Paquet Microsoft ulripa@microsoft.com Noam Koenigstein Microsoft noamko@microsoft.com |
| Pseudocode | No | The paper describes the optimization algorithm using mathematical equations and textual descriptions but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code, nor does it include links to a code repository or mention code in supplementary materials. |
| Open Datasets | Yes | Amazon Baby Registries This public dataset consists of 111,006 registries of baby products from 15 different categories... The public dataset was obtained by collecting baby registries from amazon.com and was used by previous DPP studies (Gillenwater et al. 2014; Mariet and Sra 2015). |
| Dataset Splits | Yes | To maintain consistency with prior work, we used a random split of 70% of the data for training and 30% for testing. We use a random split of 80% of the data for training and 20% for testing. We select the regularization hyperparameter, α, using a line search performed with a validation set. |
| Hardware Specification | Yes | Our implementations of the low-rank and full-rank DPP models are written in Julia, and we perform all experiments on a Windows 10 system with 32 GB of RAM and an Intel Core i7-4770 CPU @ 3.4 GHz. |
| Software Dependencies | No | The paper states 'Our implementations of the low-rank and full-rank DPP models are written in Julia' but does not provide specific version numbers for Julia or any other software dependencies. |
| Experiment Setup | Yes | In practice, we set T so that ϵ is held nearly fixed until the iteration just before the log-likelihood on a validation set begins to decrease (which indicates that we have likely jumped past the local maximum), and we find that setting β = 0.95 and ϵ0 = 1.0e-5 works well for the datasets used in this paper. We use a minibatch size of 1000 training instances, which works well for the datasets we tested. We set δ = 1.0e-5. We use K = 30 trait dimensions for the low-rank DPP models trained on this data. we use K = 15 trait dimensions for the low-rank DPP models trained on this data. |