Regularizing Towards Permutation Invariance In Recurrent Models

Authors: Edo Cohen-Karlik, Avichai Ben David, Amir Globerson

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In order to evaluate the empirical effectiveness of our regularization scheme we compare it to other methods for learning permutation invariant models. Finally, we also demonstrate how our regularization scheme is effective in semi permutation invariant settings.
Researcher Affiliation Collaboration Edo Cohen-Karlik Tel Aviv University, Israel edocohen@mail.tau.ac.il Avichai Ben David Tel Aviv University, Israel avichaib@mail.tau.ac.il Amir Globerson Tel Aviv University, Israel and Google Research gamir@post.tau.ac.il
Pseudocode No The paper describes algorithms and formulations mathematically but does not present them in structured pseudocode or algorithm blocks.
Open Source Code No The paper mentions using code from other research groups (e.g., [Murphy et al., 2018], [Lee et al., 2019]) and provides links to their implementations, but there is no explicit statement or link for the authors' own source code for the methodology described in this paper.
Open Datasets Yes We evaluate our method on a 40-way point cloud classification task using Model Net40 [Chang et al., 2015].
Dataset Splits No The paper mentions "Cross-validation was used for learning all architectures" but does not specify the type of cross-validation (e.g., k-fold) or provide exact percentages/counts for train/validation/test splits, nor does it cite specific predefined splits for reproduction.
Hardware Specification No The paper does not provide specific details about the hardware used for experiments (e.g., GPU/CPU models, memory specifications).
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions) used in their implementation.
Experiment Setup No The paper mentions 'Cross-validation was used' and 'For both networks we used the minimal width required to perfectly fit the training data', but it often defers specific hyperparameter values or detailed training configurations to other papers (e.g., 'Experiment details are as in [Zaheer et al., 2017]') or the Appendix ('For full details see the Appendix'), thus not providing concrete setup details directly in the main text for reproduction.