Papers of the day   All papers

Large Scale Adversarial Representation Learning

Comments

Jul 08 2019 DeepMind

In our new paper https://arxiv.org/abs/1907.02544 we show that GANs can be harnessed for unsupervised representation learning, with state-of-the-art results on ImageNet. Reconstructions, as shown below, tend to emphasise high-level semantics over pixel-level details. https://t.co/9gSKL6aoQu
3 replies, 953 likes


Oct 08 2019 DeepMind

We’ve released pre-trained BigBiGAN representation learning models https://arxiv.org/abs/1907.02544 on TF Hub: https://tfhub.dev/s?publisher=deepmind&q=bigbigan Try them out in a Colab at: https://colab.research.google.com/github/tensorflow/hub/blob/master/examples/colab/bigbigan_with_tf_hub.ipynb https://t.co/Hu7vPpLkgL
1 replies, 487 likes


Jul 08 2019 Andrej Karpathy

Large Scale Adversarial Representation Learning https://arxiv.org/abs/1907.02544 un/self-supervised learning is a highly fertile area (but will require much more density+structure than ImageNet affords), will obviate present necessity for datasets at scale (or rollouts in RL) #deepbelief
3 replies, 401 likes


Jul 08 2019 Ben Poole

BigBiGAN shows that "progress in image generation quality translates to substantially improved representation learning performance." Competitive w/self-supervised approaches on ImageNet. The cycle from generative models to other methods and back again continues.
4 replies, 314 likes


Jul 08 2019 Soumith Chintala

It's great to see more work on GANs focusing on downstream semi-supervised tasks, and away from pretty-picture competitions. Sometimes we lose track :)
2 replies, 191 likes


Jul 22 2019 Sam Gershman

Somebody had to do it. http://gershmanlab.webfactional.com/pubs/GenerativeAdversarialBrain.pdf
5 replies, 136 likes


Jul 08 2019 Brundage Bot

Large Scale Adversarial Representation Learning. Jeff Donahue and Karen Simonyan http://arxiv.org/abs/1907.02544
1 replies, 76 likes


Jul 09 2019 Xander Steenbrugge

This work has big implications for things like reverse image search, image clustering.. * The feature space is learned 100% unsupervised (hence it can be easily applied to any dataset) * The resulting latent space mostly cares about semantic features (which is what humans do too) https://t.co/AbtTBl6Uyu
0 replies, 34 likes


Jul 08 2019 roadrunner01

Large Scale Adversarial Representation Learning pdf: https://arxiv.org/pdf/1907.02544.pdf abs: https://arxiv.org/abs/1907.02544 https://t.co/7VE2F1aE54
0 replies, 33 likes


Jul 08 2019 Nando de Freitas

Large Scale Adversarial Representation Learning. https://arxiv.org/abs/1907.02544
0 replies, 28 likes


Jul 10 2019 Daisuke Okanohara

GAN-based representation learning (e.g., BiGAN) is considered as not effective as self-supervision representation learning such as CPC. By using BigGAN and new discriminator with unary terms for data and latent, GAN-based representation learning match CPC. https://arxiv.org/abs/1907.02544
0 replies, 7 likes


Jul 21 2019 Montreal.AI

"Large Scale Adversarial Representation Learning" Jeff Donahue and Karen Simonyan: https://arxiv.org/abs/1907.02544 #GenerativeModels #MachineLearning #RepresentationLearning https://t.co/DXts5Xpjb0
0 replies, 6 likes


Jul 08 2019 Jeffrey De Fauw

This is a great paper looking at the representation learning capabilities of GANs, which is probably a lot better than most people would have thought: 81.9 top 5 acc, 61.3 top 1 acc on ImageNet. Basically SOTA.
1 replies, 5 likes


Jul 08 2019 Elliot Turner

Unsupervised representation learning is getting real. BigBiGAN is getting 81.9% top-5 accuracy on ImageNet! Wow. https://arxiv.org/abs/1907.02544 #deepmind https://t.co/RixrNV2rQf
0 replies, 3 likes


Oct 08 2019 Brent Arnichec

We’ve released pre-trained BigBiGAN representation learning models https://arxiv.org/abs/1907.02544 on TF Hub: https://tfhub.dev/s?publisher=deepmind&q=bigbigan Try them out in a Colab at: https://colab.research.google.com/github/tensorflow/hub/blob/master/examples/colab/bigbigan_with_tf_hub.ipynb https://t.co/kJTElnszkR
0 replies, 1 likes


Jul 08 2019 Chirantan Ghosh

GANs can be harnessed for unsupervised representation learning, with state-of-the-art results on ImageNet! #gan #deeplearning #machinelearning #computervision #deepmind #imagenet #unsupervisedlearning
1 replies, 1 likes


Content