GuimPML's profile picture. Machine learning engineer at Meta, London. Sometimes I write about deep learning and GANs. Opinions are my own.

Guim Perarnau

@GuimPML

Machine learning engineer at Meta, London. Sometimes I write about deep learning and GANs. Opinions are my own.

Guim Perarnau reposted

Engineer @JoshBambrick demos NSTM: Key News Themes, a real-time query-driven news overview composition system he developed w/ @chokky_vista, @GuimPML, Igor Malioutov, @prog_aa, Vittorio Selo & Iat Chong Chan Demo 3B|8:45 AM EDT Demo 5B|4:45 PM EDT bloom.bg/3fc5DGC #ACL2020

TechAtBloomberg's tweet image. Engineer @JoshBambrick demos NSTM: Key News Themes, a real-time query-driven news overview composition system he developed w/ @chokky_vista, @GuimPML, Igor Malioutov, @prog_aa, Vittorio Selo & Iat Chong Chan
Demo 3B|8:45 AM EDT
Demo 5B|4:45 PM EDT
bloom.bg/3fc5DGC
#ACL2020

Guim Perarnau reposted

Cornell's excellent Machine Learning course is entirely online, for free, including all the lectures on YouTube and all the course notes: cs.cornell.edu/courses/cs4780… That's the bright side of the Internet: so much knowledge freely available! 🙌😀


Guim Perarnau reposted

New blog post: "A Recipe for Training Neural Networks" karpathy.github.io/2019/04/25/rec… a collection of attempted advice for training neural nets with a focus on how to structure that process over time


Guim Perarnau reposted

How do we store memories? “We’re much better at recognising than recalling. When we remember something, we have to try to relive an experience. When we recognise something, we must merely be conscious of the fact that we have had this experience before.” aeon.co/essays/your-br…

hardmaru's tweet image. How do we store memories?

“We’re much better at recognising than recalling. When we remember something, we have to try to relive an experience. When we recognise something, we must merely be conscious of the fact that we have had this experience before.”

aeon.co/essays/your-br…
hardmaru's tweet image. How do we store memories?

“We’re much better at recognising than recalling. When we remember something, we have to try to relive an experience. When we recognise something, we must merely be conscious of the fact that we have had this experience before.”

aeon.co/essays/your-br…

Guim Perarnau reposted

Large-Scale GAN Training: My internship project with Jeff and Karen. We push the SOTA Inception Score from 52 -> 166+ and give GANs the ability to trade sample variety and fidelity. arxiv.org/abs/1809.11096

ajmooch's tweet image. Large-Scale GAN Training: My internship project with Jeff and Karen.

We push the SOTA Inception Score from 52 -> 166+ and give GANs the ability to trade sample variety and fidelity.

arxiv.org/abs/1809.11096
ajmooch's tweet image. Large-Scale GAN Training: My internship project with Jeff and Karen.

We push the SOTA Inception Score from 52 -> 166+ and give GANs the ability to trade sample variety and fidelity.

arxiv.org/abs/1809.11096
ajmooch's tweet image. Large-Scale GAN Training: My internship project with Jeff and Karen.

We push the SOTA Inception Score from 52 -> 166+ and give GANs the ability to trade sample variety and fidelity.

arxiv.org/abs/1809.11096

Guim Perarnau reposted

New work from our group @GoogleAI - A practical “cookbook” for #GAN research in 2018: arxiv.org/abs/1807.04720 Code & pre-trained TensorFlow Hub models: github.com/google/compare… Attending #ICML2018 ? Visit our poster on Saturday (Reproducibility Workshop)


Guim Perarnau reposted

My new paper is out! " The relativistic discriminator: a key element missing from standard GAN" explains how most GANs are missing a key ingredient which makes them so much better and much more stable! #Deeplearning #AI ajolicoeur.wordpress.com/RelativisticGA… arxiv.org/abs/1807.00734


Guim Perarnau reposted

most common neural net mistakes: 1) you didn't try to overfit a single batch first. 2) you forgot to toggle train/eval mode for the net. 3) you forgot to .zero_grad() (in pytorch) before .backward(). 4) you passed softmaxed outputs to a loss that expects raw logits. ; others? :)


Guim Perarnau reposted

Honored to see my research together with Brian Dolhansky to be featured in #TechCrunch . Come and see our presentation at #CVPR18 next week! It will be eye opening! GANs to the power! techcrunch.com/2018/06/16/fac…

cristiancanton's tweet image. Honored to see my research together with Brian Dolhansky to be featured in #TechCrunch . Come and see our presentation at #CVPR18 next week! It will be eye opening! GANs to the power!

techcrunch.com/2018/06/16/fac…

Guim Perarnau reposted

Two years of GAN progress on class-conditional ImageNet-128

goodfellow_ian's tweet image. Two years of GAN progress on class-conditional ImageNet-128

Guim Perarnau reposted

My new paper on generating images from scene graphs using graph convolution and GANs is up on arXiv! To appear at CVPR2018, with @agrimgupta92 and @drfeifei arxiv.org/abs/1804.01622

jcjohnss's tweet image. My new paper on generating images from scene graphs using graph convolution and GANs is up on arXiv! To appear at CVPR2018, with @agrimgupta92 and @drfeifei arxiv.org/abs/1804.01622

Guim Perarnau reposted

Thread on how to review papers about generic improvements to GANs


Guim Perarnau reposted

Adversarial examples that fool both human and computer vision arxiv.org/abs/1802.08195

goodfellow_ian's tweet image. Adversarial examples that fool both human and computer vision arxiv.org/abs/1802.08195

Guim Perarnau reposted

Our new paper: "Is Generator Conditioning Causally Related to GAN Performance?" TLDR: "Almost certainly" arxiv.org/abs/1802.08768

nottombrown's tweet image. Our new paper: "Is Generator Conditioning Causally Related to GAN Performance?"

TLDR: "Almost certainly"

arxiv.org/abs/1802.08768

Guim Perarnau reposted

What's your motivation for training a generative model? Unsupervised representation learning? Computer vision? Compression? If you can answer that, you probably know what evaluation to use. Test generative models in applications, like the field used to.


Guim Perarnau reposted

Introducing exemplar GANs and a compelling use case: eye in-painting preserving the identity of the subject. #facebook #research #gan #inpainting #face #computervision arxiv.org/abs/1712.03999


Guim Perarnau reposted

An Early Overview of #ICLR2018 @ICLR2018 finally here! prlz77.github.io/iclr2018-stats Find the most exciting and the most controversial papers 😉


It has been a hot week for GANs! Here's a recap of new articles: 24 Nov - StarGANs arxiv.org/abs/1711.09020 28 Nov - Are GANs Created Equal? arxiv.org/abs/1711.10337 30 Nov - High-Resolution Image Synthesis with cGANs tcwang0509.github.io/pix2pixHD/


Loading...

Something went wrong.


Something went wrong.