Label-Aware Neural Tangent Kernel: Toward Better Generalization and Local Elasticity
Authors: Shuxiao Chen, Hangfeng He, Weijie Su
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct comprehensive experiments to confirm that our proposed LANTKs can indeed better simulate the quantitative and qualitative behaviors of NNs compared to their label-agnostic counterpart. |
| Researcher Affiliation | Academia | University of Pennsylvania {shuxiaoc@wharton, hangfeng@seas, suw@wharton}.upenn.edu |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our code is publicly available at https://github.com/Horn Hehhf/LANTK. |
| Open Datasets | Yes | On the CIFAR-10 dataset (Krizhevsky, 2009) |
| Dataset Splits | No | The paper specifies training and test data sizes and sampling constraints ('randomly sample 10000 examples as the training data and another 2000 as the test data'), but does not explicitly mention a validation set or its split. |
| Hardware Specification | No | The paper discusses computational complexity and mentions general hardware concepts (e.g., 'GPUs', 'TPUs' in references), but does not specify the exact hardware (e.g., specific GPU or CPU models, memory) used for running its own experiments. |
| Software Dependencies | No | The paper mentions that 'The implementation of CNTK is based on Novak et al. (2020)', which refers to 'Neural tangents: Fast and easy infinite neural networks in python', implying the use of Python and the Neural Tangents library, but specific version numbers for these or other software dependencies are not provided. |
| Experiment Setup | No | The paper describes general model choices and refers to appendices for architecture and method details (e.g., 'The implementation of CNTK is based on Novak et al. (2020) and the details of the architecture can be found in Appx. D.2.' and 'We refer the readers to Appx. D.3 for further details' for higher-order regression methods), but does not provide specific hyperparameter values or detailed training configurations in the main text. |