no code implementations • 11 Jan 2023 • Max Vladymyrov, Andrey Zhmoginov, Mark Sandler
We focus on the problem of learning without forgetting from multiple tasks arriving sequentially, where each task is defined using a few-shot episode of novel or already seen classes.
no code implementations • 5 Jan 2023 • Mark Sandler, Andrey Zhmoginov, Max Vladymyrov, Nolan Miller
In particular, for Exponential Moving Average (EMA) and Stochastic Weight Averaging we show that our proposed model matches the observed training trajectories on ImageNet.
1 code implementation • 15 Dec 2022 • Johannes von Oswald, Eyvind Niklasson, Ettore Randazzo, João Sacramento, Alexander Mordvintsev, Andrey Zhmoginov, Max Vladymyrov
We start by providing a simple weight construction that shows the equivalence of data transformations induced by 1) a single linear self-attention layer and by 2) gradient-descent (GD) on a regression loss.
no code implementations • CVPR 2023 • Andrey Zhmoginov, Mark Sandler, Nolan Miller, Gus Kristiansen, Max Vladymyrov
We study the effects of data and model architecture heterogeneity and the impact of the underlying communication graph topology on learning efficiency and show that our agents can significantly improve their performance compared to learning in isolation.
1 code implementation • CVPR 2022 • Mark Sandler, Andrey Zhmoginov, Max Vladymyrov, Andrew Jackson
In this paper we propose augmenting Vision Transformer models with learnable memory tokens.
1 code implementation • 11 Jan 2022 • Andrey Zhmoginov, Mark Sandler, Max Vladymyrov
In this work we propose a HyperTransformer, a Transformer-based model for supervised and semi-supervised few-shot learning that generates weights of a convolutional neural network (CNN) directly from support samples.
no code implementations • 29 Sep 2021 • Andrey Zhmoginov, Max Vladymyrov, Mark Sandler
In this work we propose a HyperTransformer, a transformer based model that generates all weights of a CNN model directly from the support samples.
no code implementations • 23 Jul 2021 • Andrey Zhmoginov, Dina Bashkirova, Mark Sandler
From practical perspective, our approach allows to: (a) reuse existing modules for learning new task by adjusting the computation order, (b) use it for unsupervised multi-source domain adaptation to illustrate that adaptation to unseen data can be achieved by only manipulating the order of pretrained modules, (c) show how our approach can be used to increase accuracy of existing architectures for image classification tasks such as ImageNet, without any parameter increase, by reusing the same block multiple times.
no code implementations • 7 May 2021 • Mingda Zhang, Chun-Te Chu, Andrey Zhmoginov, Andrew Howard, Brendan Jou, Yukun Zhu, Li Zhang, Rebecca Hwa, Adriana Kovashka
With early termination, the average cost can be further reduced to 198M MAdds while maintaining accuracy of 80. 0% on ImageNet.
Ranked #661 on Image Classification on ImageNet
1 code implementation • 10 Apr 2021 • Mark Sandler, Max Vladymyrov, Andrey Zhmoginov, Nolan Miller, Andrew Jackson, Tom Madams, Blaise Aguera y Arcas
We show that classical gradient-based backpropagation in neural networks can be seen as a special case of a two-state network where one state is used for activations and another for gradients, with update rules derived from the chain rule.
no code implementations • 10 Dec 2020 • Liangchen Luo, Mark Sandler, Zi Lin, Andrey Zhmoginov, Andrew Howard
Knowledge distillation is one of the most popular and effective techniques for knowledge transfer, model compression and semi-supervised learning.
no code implementations • 11 Aug 2020 • Mark Sandler, Andrey Zhmoginov, Liangcheng Luo, Alexander Mordvintsev, Ettore Randazzo, Blaise Agúera y Arcas
The update rule is applied repeatedly in parallel to a large random subset of cells and after convergence is used to produce segmentation masks that are then back-propagated to learn the optimal update rules using standard gradient descent methods.
no code implementations • 7 Sep 2019 • Mark Sandler, Jonathan Baccash, Andrey Zhmoginov, Andrew Howard
We explore the question of how the resolution of the input image ("input resolution") affects the performance of a neural network when compared to the resolution of the hidden layers ("internal resolution").
no code implementations • 22 Jul 2019 • Andrey Zhmoginov, Ian Fischer, Mark Sandler
We propose a new method for learning image attention masks in a semi-supervised setting based on the Information Bottleneck principle.
no code implementations • ICLR 2019 • Pramod Kaushik Mudrakarta, Mark Sandler, Andrey Zhmoginov, Andrew Howard
We introduce a novel method that enables parameter-efficient transfer and multitask learning.
no code implementations • ICLR 2019 • Pramod Kaushik Mudrakarta, Mark Sandler, Andrey Zhmoginov, Andrew Howard
We introduce a novel method that enables parameter-efficient transfer and multi-task learning with deep neural networks.
148 code implementations • CVPR 2018 • Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, Liang-Chieh Chen
In this paper we describe a new mobile architecture, MobileNetV2, that improves the state of the art performance of mobile models on multiple tasks and benchmarks as well as across a spectrum of different model sizes.
Ranked #7 on Retinal OCT Disease Classification on OCT2017
no code implementations • 8 Dec 2017 • Casey Chu, Andrey Zhmoginov, Mark Sandler
CycleGAN (Zhu et al. 2017) is one recent successful approach to learn a transformation between two image distributions.
no code implementations • 21 Feb 2017 • Soravit Changpinyo, Mark Sandler, Andrey Zhmoginov
Deep convolutional networks are well-known for their high computational and memory demands.
1 code implementation • 14 Jun 2016 • Andrey Zhmoginov, Mark Sandler
Deep neural networks have dramatically advanced the state of the art for many areas of machine learning.