FlexConv: Continuous Kernel Convolutions With Differentiable Kernel Sizes
Authors: David W. Romero, Robert-Jan Bruintjes, Jakub Mikolaj Tomczak, Erik J Bekkers, Mark Hoogendoorn, Jan van Gemert
ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate Flex Conv across classification tasks on sequential and image benchmark datasets, and validate the ability of MAGNets to approximate complex functions. A complete description of the datasets used is given in Appx. B. Appx. D.2 reports the parameters used in all our experiments.1 |
| Researcher Affiliation | Academia | 1 Vrije Universiteit Amsterdam 2 Delft University of Technology 3 University of Amsterdam The Netherlands |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found. |
| Open Source Code | Yes | Our code is publicly available at https://github.com/rjbruin/flexconv. |
| Open Datasets | Yes | All datasets used in our experiments are publicly available. |
| Dataset Splits | Yes | The Image Net-k (Chrabaszcz et al., 2017) dataset... contains 1000 classes with 1,281,167 training samples and 50,000 validation samples. |
| Hardware Specification | No | No specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) were explicitly mentioned for running the experiments. |
| Software Dependencies | No | The paper mentions using 'Weights & Biases (Biewald, 2020)' and 'torch.backends.cudnn.benchmark' but does not specify version numbers for PyTorch, CUDA, or other key software components required for replication. |
| Experiment Setup | Yes | Unless otherwise specified, we use a learning rate of 0.01 with a cosine annealing scheme (Loshchilov & Hutter, 2016) with five warmup epochs. We use a different learning rate of 0.1 the regular learning rate for the Flex Conv Gaussian mask parameters. We do not use weight decay, unless otherwise specified. |