no code implementations • 20 Mar 2023 • Erdem Koyuncu
For Rectified Linear Unit (ReLU) networks without conditional computation, it is known that memorizing a collection of $n$ input-output relationships can be accomplished via a neural network with $O(\sqrt{n})$ neurons.
no code implementations • 27 Oct 2022 • Alperen Görmez, Erdem Koyuncu
We propose Class Based Thresholding (CBT) to reduce the computational cost of early exit semantic segmentation models while preserving the mean intersection over union (mIoU) performance.
no code implementations • 8 Jul 2022 • Alperen Görmez, Erdem Koyuncu
Deep learning models that perform well often have high computational costs.
no code implementations • 10 Jun 2022 • Runxuan Miao, Erdem Koyuncu
We present federated momentum contrastive clustering (FedMCC), a learning framework that can not only extract discriminative representations over distributed local data but also perform data clustering.
no code implementations • 22 Oct 2021 • Hongyi Pan, Diaa Badawi, Runxuan Miao, Erdem Koyuncu, Ahmet Enis Cetin
In this paper, we introduce multiplication-avoiding power iteration (MAPI), which replaces the standard $\ell_2$-inner products that appear at the regular power iteration (RPI) with multiplication-free vector products which are Mercer-type kernel operations related with the $\ell_1$ norm.
no code implementations • 25 May 2021 • Hongyi Pan, Diaa Badawi, Erdem Koyuncu, A. Enis Cetin
We consider a family of vector dot products that can be implemented using sign changes and addition operations only.
1 code implementation • 1 Mar 2021 • Alperen Görmez, Venkat R. Dasari, Erdem Koyuncu
Moreover, if there are no limitations on the training time budget, E$^2$CM can be combined with an existing early exit scheme to boost the latter's performance, achieving a better trade-off between computational cost and network accuracy.
no code implementations • 23 Oct 2020 • Erdem Koyuncu
We consider quantizing an $Ld$-dimensional sample, which is obtained by concatenating $L$ vectors from datasets of $d$-dimensional vectors, to a $d$-dimensional cluster center.
no code implementations • 30 Oct 2019 • Usama Muneeb, Erdem Koyuncu, Yasaman Keshtkarjahromi, Hulya Seferoglu, Mehmet Fatih Erden, Ahmet Enis Cetin
We propose a technique to increase robustness and reduce computational complexity in a Convolutional Neural Network (CNN) based anomaly detector that utilizes the optical flow information of video data.
no code implementations • 29 Oct 2019 • Samuele Battaglino, Erdem Koyuncu
Conventional principal component analysis (PCA) finds a principal vector that maximizes the sum of second powers of principal components.