DINO as a von Mises-Fisher mixture model

Authors: Hariprasath Govindarajan, Per Sidén, Jacob Roll, Fredrik Lindsten

ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5 EXPERIMENTSWe conducted ablation experiments to study the impact of our proposed modifications to DINO.
Researcher Affiliation Collaboration 1Linköping University, Sweden 2 Qualcomm Technologies, Inc.
Pseudocode No The paper does not contain any clearly labeled pseudocode or algorithm blocks.
Open Source Code No We intend to make a public release of the code repository and pre-trained models in order to aid the research community to reproduce our experiments.
Open Datasets Yes The models are pre-trained on the Image Net dataset (Deng et al., 2009)
Dataset Splits Yes We report the k NN top-1 classification accuracy on Image Net in Table 1 by averaging over 2 runs. In Table 12 and Table 13, we show the mean and standard deviation of the top-1 validation classification accuracy over the three splits.
Hardware Specification Yes The model trainings are done on a single A100 node, consisting of 8 GPUs.
Software Dependencies No The paper mentions using the 'scipy' implementation for comparison, but no specific version numbers for any software dependencies are provided.
Experiment Setup Yes The student temperature τ is set to 0.1 and the teacher temperature is linearly scaled from 0.04 to 0.07 over some initial epochs (50 epochs for Vi T-Small/16 and 30 epochs for Vi T-Base/16). The batch sizes are adapted to fit the node and adjusted based on the model architecture (batch size=64 per GPU for Vi T-Base/16 and 128 for Vi T-Small/16).