Beyond Vector Spaces: Compact Data Representation as Differentiable Weighted Graphs
Authors: Denis Mazur, Vage Egiazarian, Stanislav Morozov, Artem Babenko
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We confirm the superiority of our method via extensive experiments on a wide range of tasks, including classification, compression, and collaborative filtering. ... Via extensive experiments on several different tasks, we confirm that, in terms of memory consumption, PRODIGE is more efficient than its vectorial counterparts. |
| Researcher Affiliation | Collaboration | Denis Mazur Yandex denismazur@yandex-team.ru Vage Egiazarian Skoltech Vage.egiazarian@skoltech.ru Stanislav Morozov Yandex Lomonosov Moscow State University stanis-morozov@yandex.ru Artem Babenko Yandex National Research University Higher School of Economics artem.babenko@phystech.edu |
| Pseudocode | No | The paper does not contain any clearly labeled pseudocode or algorithm blocks. The method is described in narrative text and mathematical formulations. |
| Open Source Code | Yes | 3. The Py Torch source code of PRODIGE is available online1. 1https://github.com/stanis-morozov/prodige |
| Open Datasets | Yes | We experiment with three publicly available datasets: MNIST10k, GLOVE10k, Celeb A10k ... All experiments are performed on the Pinterest dataset[31]. ... We evaluate our model on the IMDB benchmark [33], a popular dataset for text sentiment binary classification. |
| Dataset Splits | No | For the IMDB dataset, the paper states: "The data is split into training and test sets, each containing N=25, 000 text instances." However, it does not explicitly mention a validation set split for any of the datasets used, nor does it provide specific percentages or counts for all train/test/validation splits for all datasets. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions using "Tensor Flow or Py Torch" for autograd, "Sparse Adam" as an optimizer, "gensim model" and the "Implicit package" for baselines. However, it does not provide specific version numbers for any of these software dependencies. |
| Experiment Setup | Yes | We tune the regularization coefficient λ to achieve the overall memory consumption close to the considered operating points. ... Namely, we start with 64 edges per vertex, half of which are links to the nearest neighbors and the other half are random edges. ... we restrict a set of possible edges to include 16 user-user and item-item edges and all relevant user-item edges available in the training data; ... one-dimensional convolutional layer with 32 output filters, followed by a global max pooling layer, a Re LU nonlinearity and a final dense layer that predicts class logits. |