Ian Goodfellow
@goodfellow_ian
Followers
300,071
Following
1,116
Media
138
Statuses
2,764
Research Scientist at DeepMind. Opinions my own. Inventor of GANs. Lead author of
San Francisco, CA
Joined September 2016
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Palestina
• 261871 Tweets
Mbappe
• 261855 Tweets
Whindersson
• 246231 Tweets
Madres
• 166430 Tweets
Joost
• 154953 Tweets
Mother's Day
• 88239 Tweets
Bannon
• 74579 Tweets
الكويت
• 72461 Tweets
Felipe Neto
• 60954 Tweets
Copa América
• 50757 Tweets
#الاتحاد_الاتفاق
• 48818 Tweets
الامم المتحده
• 39739 Tweets
SALVEM O DOGUINHO DO TELHADO
• 36959 Tweets
#campusintifada
• 31502 Tweets
Is It Over Now
• 29270 Tweets
Ligue 1
• 28404 Tweets
مجلس الامه
• 27405 Tweets
Nasser
• 21737 Tweets
الجمعيه العامه
• 19748 Tweets
Rodrygo
• 19325 Tweets
Northern Lights
• 17807 Tweets
#استمرار_السالم_مرفوض3
• 17186 Tweets
#KızılcıkŞerbeti
• 16923 Tweets
$ATS
• 15732 Tweets
Jim Simons
• 11236 Tweets
صاحب السمو
• 11043 Tweets
KTLA
• 10133 Tweets
I'm excited to announce that I've joined DeepMind! I'll be a research scientist in
@OriolVinyalsML
's Deep Learning team.
152
246
7K
One of my favorite samples from the Progressive GANs paper is this one from the "cat" category. Apparently some of the cat training photos were memes with text. The GAN doesn't know what text is so it has made up new text-like imagery in the right place for a meme caption.
62
3K
6K
I never heard back from MIT. I got rejected from CMU. I was accepted to U of T but not to work with the PI I wanted there. I got "honorable mention" for NSF GRFP but not actual money. Don't let temporary failures discourage you.
46
915
4K
CycleGAN turning a horse video into a zebra video ( )
38
2K
3K
This exciting new PyTorch library includes quite a lot of the GANs I've featured in my talks over the past few years, all in one place!
10
433
2K
OctConv is a simple replacement for the traditional convolution operation that gets better accuracy with fewer FLOPs
16
572
2K
ML researchers, reviewers, and press coverage of ML need to get a lot more serious about statistically robustness of results and the effect of hyperparameters. This study shows that many papers over the last year or so were just observing sampling error, not true improvement.
33
655
2K
By looking at this image, you can see how sensitive your own eyes are to contrast at different frequencies (taller apparent peaks=more sensitivity at that frequency). It's like a graph that is made by perceiving the graph itself. h/t
@catherineols
25
529
1K
Whoa! It turns out that famous examples of NLP systems succeeding and failing were very misleading. “Man is to king as woman is to queen” only works if the model is hardcoded not to be able to say “king” for the last word.
29
386
1K
GANs for generating images of how clothes will fit. Only two of these images are photos.
19
606
1K
ML Twitter, what are your favorite papers / other resources about the class imbalance problem?
65
216
1K
GANs can be used to automatically design dental crowns, that are then actually manufactured and used in the physical world. Crowns need to be made specifically for each patient and need to fit correctly with the other teeth and support biting and chewing.
9
393
1K
Apple now has an AI/ML residency program! I'm looking forward to working with our first class of residents.
Thrilled to announce a new program designed to help experts in applied fields build ML-powered products and experiences. Introducing the AI/ML residency program:
18
208
764
11
172
1K
Philip Wang at Uber set up to show a new imaGANary person every time you refresh the page
49
430
1K
Super-resolution GANs for improving the texture resolution of old games:
21
277
1K
Adversarial examples that fool both human and computer vision
36
431
1K
Two years of GAN progress on class-conditional ImageNet-128
15
350
1K
An exciting property of style-based generators is that they have learned to do 3D viewpoint rotations around objects like cars. These kinds of meaningful latent interpolations show that the model has learned about the structure of the world.
8
356
1K
12
338
1K
The quiet semisupervised revolution continues
FixMatch: focusing on simplicity for semi-supervised learning and improving state of the art (CIFAR 94.9% with 250 labels, 88.6% with 40).
Collaboration with Kihyuk Sohn,
@chunliang_tw
@ZizhaoZhang
Nicholas Carlini
@ekindogus
@Han_Zhang_
@colinraffel
5
243
883
8
213
1K
Yoshua, Aaron, and I have released the LaTeX template for the Deep Learning book: Useful if you want to follow the same math notation conventions as we do or if you want to put a notation page in your document
8
267
1K
This new family of GAN loss functions looks promising! I'm especially excited about Fig 4-6, where we see that the new loss results in much faster learning during the first several iterations of training. I implemented the RSGAN loss on a toy problem and it worked well.
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
13
284
949
7
261
1K
I’m heading to Uruguay next month to teach about generative models at
30
83
994
Thank you to the many people who reached out after my now-deleted tweet last week asking for help with an urgent problem. For everyone still concerned, things are under control now.
38
15
939
I've spent several years studying machine learning security with the goal of making ML reliable before it is used in more and more important contexts. Unfortunately, ML capabilities and adoption are growing much faster than ML robustness.
89
164
896
Forbes listed GANs as one of the best tech innovations of the last three years:
18
193
922
MIT Technology Review listed GANs as 1 of 10 breakthrough technologies this year:
16
255
909
ML paper writing pro-tip: you can download the raw source of any arxiv paper. Click on the "Other formats" link, then click "Download source". This gets you a .tar.gz with all the .tex files, all the image files for the figures in their original resolution, etc.
19
281
908
A quick thread on two of my favorite theory hacks for machine learning research
5
253
879
While GANs have been great at generating realistic images from a single category (one GAN for faces, another GAN for buildings) they've always struggled to fit all 1,000 classes of ImageNet with a single GAN. This ICLR submission has done it: -
29
346
853
, a startup that uses GANs to generate retail marketing images, has raised $17M:
16
205
810
Self-attention for GANs. No more problems with losing track of how many faces the generator has drawn on the dog.
New preprint () by Han Zhang, with
@goodfellow_ian
and Dimitris Metaxas. Substantially improves the state-of-the-art on the conditional Imagenet synthesis task.
7
157
518
13
241
784
This is really cool. Some of my PhD labmates worked on ML for compression back in the pretraining era, and I remember it being really hard to get a compression advantage.
Check out our new work on face-vid2vid, a neural talking-head model for video conferencing that is 10x more bandwidth efficient than H264
arxiv
project
video
@tcwang0509
@arunmallya
#GAN
14
151
747
19
100
766
One of my main concerns about machine learning interpretability tools is that they will make people think they understand ML when they don't. People seem to think linear models are interpretable, but no one looks at them and has the intuition that they have adversarial examples
19
226
761
NVIDIA gave me a new T-Rex, signed by Jensen. They are not even for sale yet! Thanks NVIDIA! The pic is with GAN extraordinaire Ming-Yu at the NVIDIA reception last night.
16
49
762
To gain some idea of the far future of ML security, we studied a simple toy problem called "adversarial spheres," simulating a future where advanced ML models are extremely accurate. We find that even then, an adversary can still easily fool them.
14
259
745
If you’re upset that someone didn’t cite your paper, I strongly recommend contacting the authors privately before making a public complaint. Thread:
17
75
722
A math trick I like a lot is the approach to taking derivatives using hyperreal numbers. Thread:
15
169
714
Updating some slides from last year. “>2,000 papers later, still not really solved” -> “>5,000 papers, still not really solved”
10
47
700
Hadrien Jean has made a series of notes on linear algebra, following the deep learning textbook:
7
241
689
TensorFuzz automates the process of finding inputs that cause some specific testable behavior, like disagreement between float16 and float32 implementations of a neural network
Neural networks are notoriously hard to debug.
@gstsdn
has developed a new debugging methodology by adapting traditional coverage guided fuzzing techniques to neural networks.
3
163
505
6
232
686
#CVPR2018
I will teach an Introduction to GANs at 8:45 AM in Room 150-ABC at the Perception Beyond the Visible Spectrum workshop. Slides available at
6
166
682
When I invented adversarial training as a defense against adversarial examples, I focused on making it as cheap and scalable as possible. Eric and collaborators have now upgraded the original cheap version to compete with newer, more expensive versions.
1/ New paper on an old topic: turns out, FGSM works as well as PGD for adversarial training!*
*Just avoid catastrophic overfitting, as seen in picture
Paper:
Code:
Joint work with
@_leslierice
and
@zicokolter
to be at
#ICLR2020
3
60
238
10
114
671
I suspect that peer review *actually causes* rather than mitigates many of the “troubling trends” recently identified by
@zacharylipton
and Jacob Steinhardt:
21
216
682
The term “deep learning” reminds me of “horseless carriage.” It made sense when introduced, but now that it is the dominant paradigm, it feels quaint to specify that there is no horse. The horse here is of course the shallow model / convex cost constraint.
19
125
653
A tech report summarizing my NIPS tutorial on GANs
9
304
650
The new Justin Timberlake video is at a deep learning conference. This is just surreal to me.
29
236
640
The Self-Organizing Conference on Machine Learning is returning as a 100% online event for 2020. Nov 30-Dec 4. It will still be small to maintain the group discussion feel. Apply at
18
113
642
Thread on how to review papers about generic improvements to GANs
9
225
643
TensorFlow launches Easy Mode... I mean, Eager Mode
7
231
643
Google Brain Residency has been upgraded to Google AI Residency. Now possible to work with more AI teams at Google.
4
184
634
CycleGAN learns to turn horses into zebras *without supervision*:
12
289
610
GANs for architecture:
7
160
584
“Be careful what you wish for”
Describe programming in only six words.
We’ll RT all the best ones.
Ours:
Turning ideas and caffeine into code.
#ProgrammingIn6Words
#wednesdaywisdom
2K
376
1K
11
60
580
The "assistant professor" title seems especially galling to me: who exactly is the assistant professor assisting? They do the full job.
Academic titles tend to begin with negative adjectives and gradually remove adjectives. "Undergraduate student" -> "graduate student", "assistant professor" -> "associate professor" -> "full professor".
17
27
259
22
55
579
I think changing the name of NIPS is the right thing to do. The majority of women in the poll voted for it, and moral leadership shouldn’t be driven by polls anyway.
14
73
546
This paper shows how to make adversarial examples with GANs. No need for a norm ball constraint. They look unperturbed to a human observer but break a model trained to resist large perturbations.
7
180
534
The definition of "adversarial examples" I prefer these days is "Adversarial examples are inputs to machine learning models that an attacker has intentionally designed to cause the model to make a mistake"
15
132
517
NIPS 2017 adversarial examples challenge: Test your defenses against others' adversarial examples and vice versa!
9
215
530
David has released a new paper from an old collaboration. Glad to see it out!
2
102
514
Check out Adversarial Logit Pairing, the new state of the art defense against adversarial examples on ImageNet, by
@harinidkannan
@alexey2004
and I:
3
184
524
I'll present a talk called "Defense Against the Dark Arts" summarizing the state of the art and key research challenges for defenses against adversarial examples. Room 259, 1:30 PM.
9
153
517
A great Scientific American article on creativity and problem solving in machine learning by Chris Baraniuk:
3
210
513
Deep learning for predicting aftershocks of large earthquakes. Besides offering better predictions, interpretations of the model suggest promising directions for new physical theories
4
170
509
It’s strange to see people defining deep learning as supervised learning via backprop, considering that the 2006 deep learning revolution was originally based on the idea that neither of those things work very well
10
112
496
I originally thought of GANs as an unsupervised learning algorithm, but so far, to create recognizable object categories, they've needed a supervision signal / labeled images. This new work shows how to get them to work well with few labels.
How to train SOTA high-fidelity conditional GANs usin 10x fewer labels? Using self-supervision and semi-supervision! Check out our latest work at
@GoogleAI
@ETHZurich
@TheMarvinRitter
@mtschannen
@XiaohuaZhai
@OlivierBachem
@sylvain_gelly
1
69
243
3
138
502
Neural networks are notoriously hard to debug.
@gstsdn
has developed a new debugging methodology by adapting traditional coverage guided fuzzing techniques to neural networks.
3
163
505
The arxiv of the future must have comments and open peer review:
17
176
501
Colin was a senior research scientist in my team at Google. He's done great technical work, especially on attention models and semi-supervised / transfer learning, and has been an excellent mentor for many Brain residents / interns. Will definitely be a great PhD advisor.
I'm starting a professorship in the CS department at UNC in fall 2020 (!!) and am hiring students! If you're interested in doing a PhD
@unccs
please get in touch. More info here:
82
146
893
2
34
498
Updates about SOCML: 1) I have failed to run a SOCML 2019 2) I’m not quitting, just having a busy year. I intend to run SOCML 2020 and beyond 3) We’re experimenting with a distributed SOCMLx program. See link for details
39
52
491
StarGAN: learning one model that translates between *multiple* domains without supervision (previous works were about translating between two domains without supervision)
0
186
483
#CVPR2018
Check out the all-day tutorial on GANs tomorrow: I'll speak at 9AM giving an introduction to GANs. Many great speakers throughout the day!
9
138
473
Interpretation of a machine learning model by a human involves both the model and the human. Human misconceptions can cause as much trouble as any property of the model.
Remarkable finding: people don't trust transparent models any more than opaque ones, and have more difficulty detecting large errors in transparent ones:
17
229
559
6
132
473
Our work on adversarial examples for the human vision system has been accepted to
#COSYNE19
:
6
83
465
Interested in jump-starting your career in machine learning research? Consider the Google AI Residency Program! Applications are now open until January 28th, 2019! Check out for more information.
5
172
462
Apple has launched an online machine learning journal, similar to Distill. First article is on GANs!
3
184
461
Differentially private GANs generate fake clinical trial data to train any ML algorithm without sacrificing privacy:
2
241
449
It’s interesting to see the pendulum swing back to representation learning. During my PhD, most of my collaborators and I were primarily interested in representation learning as a biproduct of sample generation, not sample generation itself.
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.
3
66
298
5
63
446
Check out
@fermatslibrary
's Librarian, a Chrome extension that automatically shows comments for ArXiv papers: I've asked for this feature for a long time!
5
162
442
In our previous collaboration's I've benefited a lot from using David's machine learning frameworks. If you're a researcher or student looking for both simplicity and customizability, definitely check out Objax.
I’d like to share my new project: Objax, a new high-level JAX API with a PyTorch-like interface! Objax pursues the quest for the simplest design and code that’s as easy as possible to extend without sacrificing performance.
1/7
10
135
555
3
71
435
The discriminator often knows something about the data distribution that the generator didn't manage to capture. By using rejection sampling, it's possible to knock out a lot of bad samples.
New preprint by
@smnh_azadi
,
@catherineols
, Trevor Darrell,
@goodfellow_ian
, and me: . We perform rejection sampling on a trained GAN generator using a GAN discriminator. This helps quite a lot for not-much effort.
1
65
231
2
104
438