Abstract | ||
---|---|---|
Convolutional layers are one of the basic building blocks of modern deep neural networks. One fundamental assumption is that convolutional kernels should be shared for all examples in a dataset. We propose conditionally parameterized convolutions (CondConv), which learn specialized convolutional kernels for each example. Replacing normal convolutions with CondConv enables us to increase the size and capacity of a network, while maintaining efficient inference. We demonstrate that scaling networks with CondConv improves the performance and inference cost trade-off of several existing convolutional neural network architectures on both classification and detection tasks. On ImageNet classification, our CondConv approach applied to EfficientNet-B0 achieves state-of-the-art performance of 78.3% accuracy with only 413M multiply-adds. Code and checkpoints for the CondConv Tensorflow layer and CondConv-EfficientNet models are available at: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet/condconv. |
Year | Venue | Field |
---|---|---|
2019 | ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 32 (NIPS 2019) | Parameterized complexity,Mathematical optimization,Inference,Convolution,Computer science,Theoretical computer science |
DocType | Volume | ISSN |
Conference | 32 | 1049-5258 |
Citations | PageRank | References |
1 | 0.35 | 0 |
Authors | ||
4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Brandon Yang | 1 | 1 | 0.69 |
Gabriel Bender | 2 | 157 | 8.35 |
Quoc V. Le | 3 | 8501 | 366.59 |
Jiquan Ngiam | 4 | 297 | 19.56 |