Filters








4 Hits in 1.8 sec

NetTailor: Tuning the Architecture, Not Just the Weights

Pedro Morgado, Nuno Vasconcelos
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
In this way, NETTAILOR can adapt the network architecture, not just its weights, to the target task.  ...  Under the standard paradigm of network fine-tuning, an entirely new CNN is learned per task, and the final network size is independent of task complexity.  ...  NETTAILOR is a flexible transfer procedure that adjusts the network architecture (not just the weights) while keeping the majority of the parameters unchanged.  ... 
doi:10.1109/cvpr.2019.00316 dblp:conf/cvpr/MorgadoV19 fatcat:a3j3olxw2vcobmjb2jfvs64zty

Incremental multi-domain learning with network latent tensor factorization [article]

Adrian Bulat and Jean Kossaifi and Georgios Tzimiropoulos and Maja Pantic
2019 arXiv   pre-print
We do so by jointly parametrizing weights across layers using low-rank Tucker structure. The core is task agnostic while a set of task specific factors are learnt on each new domain.  ...  the training sets.  ...  Our approach is in particular not architecture specific and can be used for various network architectures.  ... 
arXiv:1904.06345v2 fatcat:5mjv6wfxubfethf3644nzqjw6i

Incremental Multi-Domain Learning with Network Latent Tensor Factorization

Adrian Bulat, Jean Kossaifi, Georgios Tzimiropoulos, Maja Pantic
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
We do so by jointly parametrizing weights across layers using low-rank Tucker structure. The core is task agnostic while a set of task specific factors are learnt on each new domain.  ...  the training sets.  ...  Our approach is in particular not architecture specific and can be used for various network architectures.  ... 
doi:10.1609/aaai.v34i07.6617 fatcat:mymyvvo4h5bcxg6avnocfxpt6y

Shape Adaptor: A Learnable Resizing Module [article]

Shikun Liu, Zhe Lin, Yilin Wang, Jianming Zhang, Federico Perazzi, Edward Johns
2020 arXiv   pre-print
The source code is available at: https://github.com/lorenmt/shape-adaptor.  ...  Our implementation enables shape adaptors to be trained end-to-end without any additional supervision, through which network architectures can be optimised for each individual task, in a fully automated  ...  Acknowledgements We would like to thank the helpful discussions from Linxi Fan and Bingyu Zhang.  ... 
arXiv:2008.00892v2 fatcat:v2hto5hlzzfnvow6slsl2m5tuy