Pseudo-spherical Knowledge Distillation

Authors: Kyungmin Lee, Hyeongkeun Lee

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through experiments on various model compression tasks, we validate the effectiveness of our method by showing its superiority over the original knowledge distillation.
Researcher Affiliation Academia Kyungmin Lee , Hyeongkeun Lee Agency for Defense Development {kyungmnlee, lhk528}@gmail.com
Pseudocode No No pseudocode or algorithm blocks were found in the paper.
Open Source Code No The paper mentions using pre-trained teacher models from an official CRD repository (https://github.com/Hobbit Long/Rep Distiller), but does not explicitly state that the code for their own method (PSKD) is open-source or provide a link to it.
Open Datasets Yes Through experiments on CIFAR-100 and Image Net model compression benchmarks, we demonstrate the effectiveness of our new distillation methods.
Dataset Splits Yes Table 4: Top-1 and Top-5 error rates (%) of student network Res Net-18 on Image Net validation set.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory) used for experiments are mentioned in the paper.
Software Dependencies No The paper mentions using 'official Py Torch pretrained Res Net-34' but does not specify the version of PyTorch or any other software dependencies with version numbers.
Experiment Setup No The paper mentions following model compression benchmarks from CRD [Tian et al., 2019] and using balancing weights α, β, η > 0, and specific γ values (0.5, 1.0), but does not provide explicit details like learning rates, batch sizes, number of epochs, or optimizer settings for the experimental setup.