On the Universality of Rotation Equivariant Point Cloud Networks
Authors: Nadav Dym, Haggai Maron
ICLR 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | H EXPERIMENTS Table 1: Results obtained on the QM9 dataset Ramakrishnan et al. (2014) for different design choices in the TFN architecture. Results are reported for the ϵhomo target variable, and are multiplied by 103. Model variant Mean ℓ1 error Order 0 irreps 111.2 ± 0.3 Order 0-1 irreps 82.1 ± 0.4 Order 0-2 irreps 62.1 ± 1.5 Order 0-3 irreps 53.2 ± 1.9 Order 0-4 irreps 51.2 ± 0.4 Order 0-3 irreps without non-linearity 53.1 ± 0.8 Order 0-3 irreps + self-interaction only in the final layer 55.2 ± 0.3 |
| Researcher Affiliation | Collaboration | Nadav Dym Duke University nadavdym@gmail.com Haggai Maron NVIDIA Research hmaron@nvidia.com |
| Pseudocode | No | The paper does not contain any sections or figures explicitly labeled 'Pseudocode' or 'Algorithm'. |
| Open Source Code | No | The paper mentions 'We used the TFN implementation from Fuchs et al. (2020)' and acknowledges Fabian B. Fuchs for 'making code available', but it does not state that the code for *this* paper's specific methodology is open source or provide a direct link to it. |
| Open Datasets | Yes | We use the QM9 (Ramakrishnan et al., 2014) dataset for our experiments. |
| Dataset Splits | Yes | We used the default parameters and data splits from Fuchs et al. (2020). |
| Hardware Specification | Yes | All experiments ran on NVIDIA GV100 GPUs. |
| Software Dependencies | No | The paper mentions 'pytorch (Paszke et al., 2017)' and 'Deep Graph Library (DGL) (Wang et al., 2019a)' but does not provide specific version numbers for these software dependencies. |
| Experiment Setup | Yes | We trained each model variant for 50 epochs on the ϵhomo target variable using an ℓ1 loss function and the ADAM optimizer with learning rate 10−3 and report results on the test set on the final epoch averaged over two runs. |