Thomas Kipf @tkipf Twitter profile | Pikagi

Pikagi

Thomas Kipf

@tkipf

25,502

Followers

1,225

Following

220

Media

1,598

Statuses

AI Research at @GoogleDeepMind . Ex-Physicist. Graph Neural Networks & Controllable Generative Models (e.g. GCNs, Structured World Models, Slot Attention).

SF Bay Area

https://t.co/aonDi5HU5B

Joined June 2009

Don't wanna be here? Send us removal request.

Pinned Tweet

@tkipf

Thomas Kipf

4 years

My PhD thesis "Deep Learning with Graph-Structured Representations" is now available for download: -- It covers a range of emerging topics in Deep Learning: from graph neural nets (and graph convolutions) to structure discovery (objects, relations, events)

Tweet media one

39

631

3K

Last Seen Profiles

@stw_pdg

@IceAndFireOmid

@stw46

@galery_basah10

@LittleCryingWey

@theimpulsivebuy

@geoemendezz

@gatesfoundation

@sorong_lion

@Fvsari

@danmckiedj

@futbol_fantasy

@BrawlStars

@vandacelia

@_JadaMC_

@Avalix

@PrinGavrilo

@huanghuO_o

@stw_pdg

@gxigi_

@pengen_stw

@Lil_Patrickk

@jandakembangstw

@MariaRamos69977

@StalkingVi96051

@LittleCryingWey

@CEDOWN

@andrewvoerman

@WaguiSpace

@jandakembangstw

@NSepehrvand

@Cs_Styx

@GoogleArabia

@ilyassahinMD

@smolsungies

@pengen_stw

@tkipf

Thomas Kipf

4 years

Very excited about the release of Jraph. Finally an easy-to-use, extensive and fast Graph Neural Network library in JAX! Fully compatible with NN libraries such as Flax and Haiku:

Tweet media one

14

178

1K

@tkipf

Thomas Kipf

6 years

DeepMind is releasing their GraphNets library: - a very comprehensive and easy-to-use library for training graph (neural) networks and related models

Tweet media one

8

487

1K

@tkipf

Thomas Kipf

4 years

I’m very excited to announce that I have joined the Google Brain team in Amsterdam 🚲🚣‍♀️🌳 as a Research Scientist! Looking fwd to both continuing my focus on graph-structured representation learning and to explore novel directions with an amazing team of new and old colleagues 😊

46

30

1K

@tkipf

Thomas Kipf

4 years

Excited to share our work @GoogleAI on Object-centric Learning with Slot Attention! Slot Attention is a simple module for structure discovery and set prediction: it uses iterative attention to group perceptual inputs into a set of slots. Paper: [1/7]

Tweet media one

12

174

712

@tkipf

Thomas Kipf

2 years

So excited to share Object Scene Representation Transformer (OSRT): OSRT learns about complex 3D scenes & decomposes them into objects w/o supervision, while rendering novel views up to 3000x faster than prior methods! 🖥️ 📜 1/7

5

117

646

@tkipf

Thomas Kipf

4 years

Thanks, @wellingmax , my committee, and everyone who accompanied me through this journey in the last four years!

@wellingmax

Max Welling

4 years

Congratulations to dr. Thomas Kipf who just successfully defended his thesis through zoom. He graduated CUM LAUDE!!

Tweet media one

14

21

707

56

11

630

@tkipf

Thomas Kipf

5 years

Excited to share our work on Contrastive Learning of Structured World Models! C-SWMs learn object-factorized models & discover objects without supervision, using a simple loss inspired by work on graph embeddings Paper: Code: 1/5

Tweet media one

5

146

601

@tkipf

Thomas Kipf

5 years

Interested in the latest developments related to Graph Neural Nets and Structured Deep Learning? The talk recordings from the @icmlconf workshop on Learning and Reasoning with Graph-Structured Representations are now available:

Tweet media one

3

138

530

@tkipf

Thomas Kipf

8 years

We are releasing the code for Semi-Supervised Classification with Graph Convolutional Networks (in TensorFlow):

Tweet media one

5

276

520

@tkipf

Thomas Kipf

3 years

Very honored to receive the ELLIS PhD Award for my thesis on Deep Learning with Graph-Structured Representations -- alongside with with @NagraniArsha for her work on multimodal DL (congrats!)

Tweet media one

@fabian_theis

Fabian Theis

3 years

@ELLISforEurope General Assembly happening just now, with ELLIS PhD Award 2021 to @NagraniArsha and @thomaskipf for very impressive theses on multimodal deep learning and graph NNs, respectively - congrats to both!

Tweet media one

2

8

80

38

34

508

@tkipf

Thomas Kipf

9 months

Life update: I've moved to the SF Bay Area! Excited to work more closely with my US-based @GoogleDeepMind colleagues and to meet both old and new friends in the area. Leaving Amsterdam wasn't an easy decision: it's such an amazing city with a vibrant ML/AI community. 1/3

Tweet media one

23

5

508

@tkipf

Thomas Kipf

7 years

Our implementation of graph auto-encoders (in TensorFlow) is now available on GitHub:

Tweet media one

3

213

496

@tkipf

Thomas Kipf

6 years

By far the cleanest and most elegant library for graph neural networks in PyTorch. Highly recommended! Unifies Capsule Nets (GNNs on bipartite graphs) and Transformers (GCNs with attention on fully-connected graphs) in a single API.

@PyTorch

PyTorch

6 years

DGL (Deep Graph Library) - Clean and efficient library to build graph neural networks including GCN, TreeLSTM and graph generative models. Includes auto-batching and other tricks for speed.

Tweet media one

4

280

768

1

130

451

@tkipf

Thomas Kipf

7 years

PyTorch implementation of Graph Convolutional Networks (with sparse mx support) - faster than in TensorFlow on GPU

Tweet media one

1

162

442

@tkipf

Thomas Kipf

7 months

"We [...] empirically show that explicitly enforcing roto-translation equivariance is not a strong requirement for generalization." "Furthermore, we also show that approaches that do not explicitly enforce roto-translation equivariance (like ours) can match or outperform

Tweet media one

@YuyangW95

Yuyang Wang

7 months

1/n New preprint alert! Introducing Generative Molecular Conformer Fields (MCF) a generative model for molecular conformer generation that obtains state-of-the-art results without using any domain specific inductive biases!

4

33

246

8

45

431

@tkipf

Thomas Kipf

6 years

We have released the code (in keras/TensorFlow) for our ESWC 2018 paper: Modeling Relational Data with Graph Convolutional Networks -

Tweet media one

1

137

408

@tkipf

Thomas Kipf

5 years

Slides and recording of my talk/tutorial on Unsupervised Learning with Graph Neural Networks at the @ipam_ucla Workshop on Deep Geometric Learning of Big Data and Applications are now available: Slides: Recording:

Tweet media one

1

102

404

@tkipf

Thomas Kipf

6 years

Happy to share our new work on MolGAN: An implicit generative model for small molecular graphs. GANs + RL for generating graphs with desired properties: - Some benefits over VAEs, but can still suffer from mode collapse. Joint work with @nicola_decao

Tweet media one

Tweet media two

Tweet media three

5

158

400

@tkipf

Thomas Kipf

7 years

Re-implemented graph convolutional nets in PyTorch as an afternoon project today: massive speed-up (~5x on Cora) compared to TensorFlow!

Tweet media one

9

99

384

@tkipf

Thomas Kipf

4 years

We have released the code for Slot Attention (incl. pre-trained model checkpoints on CLEVR) Code: NeurIPS camera ready:

Tweet media one

1

58

376

@tkipf

Thomas Kipf

2 years

Happy to announce that there will be a workshop on Neuro Causal & Symbolic AI (nCSI) at @NeurIPSConf ! More information soon. 🖥️

Tweet media one

6

57

369

@tkipf

Thomas Kipf

4 years

Happy to learn that our work on Slot Attention has been accepted for spotlight presentation at @NeurIPSConf !

@tkipf

Thomas Kipf

4 years

Excited to share our work @GoogleAI on Object-centric Learning with Slot Attention! Slot Attention is a simple module for structure discovery and set prediction: it uses iterative attention to group perceptual inputs into a set of slots. Paper: [1/7]

Tweet media one

12

174

712

4

42

368

@tkipf

Thomas Kipf

6 years

We just released a TensorFlow implementation of our Graph Convolutional Matrix Completion paper: (with @vdbergrianne ) - A recommender system based on graph neural networks

Tweet media one

2

125

367

@tkipf

Thomas Kipf

3 years

Excited to share our work on Conditional Object-Centric Learning from Video! We introduce SAVi, a slot-based model that can discover + represent visual entities in videos, using simple location cues and object motion (...or entirely unsupervised) 🖥️ 1/7

5

61

360

@tkipf

Thomas Kipf

2 months

👀

Tweet media one

@GoogleDeepMind

Google DeepMind

@GoogleDeepMind

2 months

Announcing AlphaFold 3: our state-of-the-art AI model for predicting the structure and interactions of all life’s molecules. 🧬 Here’s how we built it with @IsomorphicLabs and what it means for biology. 🧵

216

2K

7K

6

22

364

@tkipf

Thomas Kipf

5 years

PyTorch Geometric has been growing into a fantastic library for graph neural nets and related methods. Great work by @rusty1s and Jan Eric Jenssen-

Tweet card media

GitHub - pyg-team/pytorch_geometric: Graph Neural Network Library for PyTorch

Graph Neural Network Library for PyTorch. Contribute to pyg-team/pytorch_geometric development by creating an account on GitHub.

@StatMLPapers

Stat.ML Papers

5 years

Fast Graph Representation Learning with PyTorch Geometric. (arXiv:1903.02428v1 [cs.LG])

0

17

82

5

74

341

@tkipf

Thomas Kipf

3 years

To showcase best practices for building/training Graph Neural Nets in JAX, we put together a comprehensive example for molecular activity prediction using Flax & Jraph Official Flax GNN example: Great work by @BigAmeya w/ collaborators @ Brain & DeepMind

Tweet media one

@BigAmeya

Ameya Daigavane

3 years

The official 'Graph Neural Networks for Molecular Activity Predictions' example for Flax is now up! (1/n)

Tweet media one

3

43

221

1

65

341

@tkipf

Thomas Kipf

6 years

"Neural Relational Inference for Interacting Systems" is accepted to @icmlconf (w/ @EthanFetaya , Jackson Wang, @wellingmax , Rich Zemel)! Graph neural nets can learn to infer latent relational structure and model multi-agent and physical dynamics 👉

Tweet media one

Tweet media two

Tweet media three

3

93

338

@tkipf

Thomas Kipf

2 years

A beautifully illustrated introduction to graph neural networks!

@rishabh16_

Rishabh Anand 🧬

3 years

🚨NEW WEEKEND READING🚨 I'm publishing "Graph Neural Networks for Novice Math Fanatics" – a primer on the math behind GNNs using colourful drawings & diagrams (I hope this becomes the definitive guide to GDL for those hoping to enter the field!!!)

Tweet media one

Tweet media two

Tweet media three

Tweet media four

10

232

1K

1

60

326

@tkipf

Thomas Kipf

6 years

PinSage: A New Graph Convolutional Neural Network for Web-Scale Recommender Systems - GCNs for network data made it from small academic benchmarks () to a web-scale industry application in less than two years time

Tweet card media

PinSage: A new graph convolutional neural network for web-scale recommender systems

Ruining He | Pinterest engineer, Pinterest Labs

1

105

320

@tkipf

Thomas Kipf

6 years

Large-scale Graph Neural Network training without subsampling! Ma et al. ( @MSFTResearch ) implement a "Gather-Apply-Scatter" (GAS) approach in TensorFlow for Multi-GPU, distributed training of GNNs/GCNs:

Tweet media one

1

71

317

@tkipf

Thomas Kipf

2 years

The submission portal for the @iclr_conf Workshop on the Elements of Reasoning: Objects, Structure, and Causality (OSC) will open next Monday! #osc2022 Submit your work by 25 Feb (end of day, AoE) 👉

1

52

318

@tkipf

Thomas Kipf

6 years

Our implementation (in PyTorch) of "Neural Relational Inference for Interacting Systems" is available on GitHub: - includes a variant of Graph Neural Networks that learns both node- and edge-based representations. Joint work with @EthanFetaya

Tweet media one

0

108

304

@tkipf

Thomas Kipf

4 years

Video recordings of (most) invited talks are now available on the workshop website: incl. talks by @PeterWBattaglia , @NataliaNeverova , @lipmanya , Qi Liu, @miltos1 , Noemi Montobbio & @pimdehaan

@tkipf

Thomas Kipf

4 years

We are organizing an @ELLISforEurope Workshop on Geometric and Relational Deep Learning! Registration invites will be shared soon. Interested in participating? Consider submitting an abstract or get in touch: w/ @erikjbekkers @wellingmax @mmbronstein

Tweet media one

3

81

279

2

67

284

@tkipf

Thomas Kipf

6 years

Zero-shot Recognition via Semantic Embeddings and Knowledge Graphs (CVPR2018): Graph Convolutional Networks (GCNs) for zero-shot learning - massive improvements on a challenging ImageNet task

Tweet media one

0

86

283

@tkipf

Thomas Kipf

4 years

We are organizing an @ELLISforEurope Workshop on Geometric and Relational Deep Learning! Registration invites will be shared soon. Interested in participating? Consider submitting an abstract or get in touch: w/ @erikjbekkers @wellingmax @mmbronstein

Tweet media one

3

81

279

@tkipf

Thomas Kipf

5 years

Our paper on Structured World Models got accepted to #ICLR2020 as a long talk! @iclr_conf Congrats to amazing co-authors @ElisevanderPol @wellingmax

Tweet card media

Contrastive Learning of Structured World Models

Contrastively-trained Structured World Models (C-SWMs) learn object-oriented state representations and a relational model of an environment from raw pixel input.

@tkipf

Thomas Kipf

5 years

Excited to share our work on Contrastive Learning of Structured World Models! C-SWMs learn object-factorized models & discover objects without supervision, using a simple loss inspired by work on graph embeddings Paper: Code: 1/5

Tweet media one

5

146

601

5

42

275

@tkipf

Thomas Kipf

1 year

Excited to announce DORSal: a 3D structured diffusion model for generation and object-level editing of 3D scenes. DORSal is “geometry-free” and learns 3D scene structure purely from data – no expensive volume rendering! 🖥️ 📜 1/6

3

66

274

@tkipf

Thomas Kipf

6 years

Slides from my talk at University of Cambridge from last Friday are now available. Thanks to @PetarV_93 for the invitation! Link: (warning: 15MB!)

Tweet media one

4

75

262

@tkipf

Thomas Kipf

6 years

Just finished reading through DeepMind's / @PeterWBattaglia et al.'s new review/position paper on graph neural nets: Timely and highly relevant contribution to the field IMO, couldn't agree more with their motivation for this class of models

Tweet card media

Relational inductive biases, deep learning, and graph networks

Artificial intelligence (AI) has undergone a renaissance recently, making major progress in key domains such as vision, language, control, and decision-making. This has been due, in part, to cheap...

4

69

259

@tkipf

Thomas Kipf

6 years

Building Models that Learn to Discover Structure and Relations - A short post about our @icmlconf paper on Neural Relational Inference. w/ @EthanFetaya

Tweet card media

Building Models that Learn to Discover Structure and Relations

Thomas Kipf, Ethan Fetaya, Jackson Wang, Max Welling, Rich Zemel (ML Research Partners from University of Amsterdam)

1

73

261

@tkipf

Thomas Kipf

6 years

Our latest take on graph neural nets: inferring latent interaction graphs - Neural Relational Inference for Interacting Systems, with @EthanFetaya

Tweet media one

0

84

256

@tkipf

Thomas Kipf

5 years

Camera-ready versions of accepted papers at the @NeurIPSConf Graph Representation Learning Workshop 2019 are now online! Some are missing, since camera-ready submission was optional (due to "no pre-print" policy of some other conferences/journals).

Tweet media one

1

71

256

@tkipf

Thomas Kipf

6 years

Our work on Hyperspherical Variational Auto-Encoders is accepted at @uai2018 (plenary talk)! Great work from MSc students @im_td @lcfalors @nicola_decao (with @jmtomczak ). Time to swap out your dusty Euclidean latent spaces 💨

1

61

256

@tkipf

Thomas Kipf

2 years

Amsterdam is taking city infrastructure to the next level: can't stress enough how amazing these new, incredibly spacious (and surprisingly red) bike paths are.

Tweet media one

8

6

249

@tkipf

Thomas Kipf

5 years

Accepted papers at the @icmlconf Workshop on Learning and Reasoning with Graph-Structured Data are now available on the workshop website: Papers: Schedule:

Tweet media one

0

60

253

@tkipf

Thomas Kipf

3 years

Exciting news: @iclr_conf will be hosting a Workshop on the "Elements of Reasoning: Objects, Structure and Causality" -- more information soon!

4

28

253

@tkipf

Thomas Kipf

2 years

Blog post about how Airbnb is using Graph Convolutional Networks (and more recent GCN variants) in production:

Tweet card media

Graph Machine Learning at Airbnb

How Airbnb is leveraging graph neural networks to up-level our machine learning

2

41

250

@tkipf

Thomas Kipf

8 years

My first-ever blog post is now online: Graph Convolutional Networks - An introduction to neural networks on graphs

2

91

243

@tkipf

Thomas Kipf

6 years

New paper on learning hyperspherical latent spaces: Hyperspherical Variational Auto-Encoders (with @im_td , L. Falorsi, @nicola_decao , @jmtomczak ). Useful trick for learning node embeddings in graphs and for semi-supervised learning.

Tweet media one

3

72

238

@tkipf

Thomas Kipf

14 days

Excited to share our work on Neural Assets: a new method for enabling 3D asset-level control in image diffusion models – scalable & without any 3D inductive biases. Neural Assets goes beyond text or pixel-based control & provides an interface inspired by 3D graphics tools. 🧵

3

45

244

@tkipf

Thomas Kipf

3 years

One year ago today, I defended my PhD thesis (). While the past year certainly has felt like any kind of celebration is inappropriate, I am still deeply grateful for everyone who helped or accompanied me along this journey.

Tweet media one

3

17

235

@tkipf

Thomas Kipf

2 years

Accepted papers are now available on the ICLR Objects, Structure & Causality (OSC) workshop website! 👉 We have received a total of 42 submissions out of which 31 will be presented at the workshop on April 29

4

39

238

@tkipf

Thomas Kipf

4 years

The (virtual) ELLIS Workshop on Geometric and Relational Deep Learning will happen this Friday! The invited talks will be live-streamed on YouTube:

@tkipf

Thomas Kipf

4 years

We are organizing an @ELLISforEurope Workshop on Geometric and Relational Deep Learning! Registration invites will be shared soon. Interested in participating? Consider submitting an abstract or get in touch: w/ @erikjbekkers @wellingmax @mmbronstein

Tweet media one

3

81

279

4

77

232

@tkipf

Thomas Kipf

4 years

Internship applications for Google Research are now open. If you are interested in working with some of us in the Brain Team in Europe (or elsewhere), please apply! Note that all internships will be virtual again next year.

Search Jobs - Google Careers

Find your next job at Google — Careers at Google. Search by location, role, skills, and more.

4

48

229

@tkipf

Thomas Kipf

4 years

It's exciting to see that Neural Relational Inference () can be re-purposed as a state-of-the-art causal discovery method (in a sense of non-linear Granger causality), entirely based on graph neural networks and variational inference.

Tweet card media

Neural Relational Inference for Interacting Systems

Interacting systems are prevalent in nature, from dynamical systems in physics to complex societal dynamics. The interplay of components can give rise to complex behavior, which can often be...

@david_madras

David Madras

4 years

New pre-print up with @sindy_loewe , Rich Zemel, and @wellingmax about causal discovery in time-series data! "Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Data": THREAD BELOW ⬇️⬇️⬇️

4

45

185

1

34

220

@tkipf

Thomas Kipf

1 year

The recording of our @CVPR Tutorial on Object Localization for Free is now available! 📽️ 🖥️

Tweet media one

1

55

218

@tkipf

Thomas Kipf

5 years

The list of accepted papers at the @NeurIPSConf Graph Representation Learning Workshop 2019 is online! (Camera-ready versions will follow later this month). Submission statistics / acceptance rates below 👇

Accepted Papers

NeurIPS 2019 Workshop

grlearning.github.io

2

55

214

@tkipf

Thomas Kipf

5 years

Excited to share @davide__belli 's work (with minor contributions by myself) on the Generative Graph Transformer, to be presented at the Graph Representation Learning workshop at @NeurIPSConf . Blog: Paper: 1/2

Tweet media one

1

45

209

@tkipf

Thomas Kipf

2 years

Graph Neural Networks for System Interaction Inference: A very nice article/tutorial by @MJAljubran on how to use Neural Relational Inference to infer hidden relations in interacting systems using GNNs 👉

0

30

207

@tkipf

Thomas Kipf

3 years

SLATE (Slot Attention Transformer) is a new approach for unsupervised learning of compositional abstractions from visual input. Combines Slot Attention w/ DALL-E to 1) bring object compositionality to DALL-E (in the absence of text prompts) and ... 1/2

@SungjinAhn_

Sungjin Ahn

3 years

Check out our new paper: "Illiterate DALL.E learns to compose." We propose a slot-based autoencoding arch, called SLATE, for learning object-centric representations that allow systematic generalization in zero-shot image generation without text!

2

29

115

2

39

202

@tkipf

Thomas Kipf

4 years

SCOUTER: An explainable image classifier using a modified version of Slot Attention by Liangzhi Li et al. (Osaka University) SCOUTER: Slot Attention:

Tweet media one

Tweet media two

2

42

199

@tkipf

Thomas Kipf

7 years

Graph neural nets make their way into computer vision: - learn interaction graph (data-dependent) and propagate graph NN-style messages. Cool application!

Tweet media one

1

66

190

@tkipf

Thomas Kipf

5 years

The call for papers for our @NeurIPSConf Graph Representation Learning Workshop is out! Submit your papers by 9 September: w/ co-organizers @vdbergrianne , @mmbronstein , @williamleif , Stefanie Jegelka, @jure , @lrjconan , @yizhousun , @PetarV_93

Tweet media one

1

48

193

@tkipf

Thomas Kipf

5 years

We are taking in submissions for the Representation Learning on Graphs and Manifolds workshop at @iclr2019 . Deadline: 22 March. Submission details:

Tweet media one

1

51

190

@tkipf

Thomas Kipf

7 years

Recommendation with (convolutional) graph auto-encoders - joint work with @vdbergrianne

Tweet media one

1

70

186

@tkipf

Thomas Kipf

2 years

Excited to share our work on self-supervised video object representation learning: We introduce SAVi++, a slot-based video model that — for the first time — scales to Waymo Open driving scenes w/o direct supervision. 🖥️ 📜 1/7

3

35

180

@tkipf

Thomas Kipf

8 years

Interested in deep learning on graphs? Come by my poster on VAEs for graphs at the Bayesian Deep Learning w/s and say hi! #nips2016

Tweet media one

4

79

174

@tkipf

Thomas Kipf

3 years

Excited (and honored) to be elected as an @ELLISforEurope Scholar in the newly launched "Semantic, Symbolic and Interpretable Machine Learning" ELLIS program. Many thanks to @vtresp , @kerstingAIML & Paolo Frasconi for the nomination and to many others in ELLIS for their support!

Tweet media one

9

6

171

@tkipf

Thomas Kipf

5 years

Very happy to learn that our workshop proposal on Graph Representation Learning at @NeurIPSConf was accepted! w/ great co-organizers (see below) + @vdbergrianne , @lrjconan , Stefanie Jegelka and Yizhou Sun

@mmbronstein

Michael Bronstein

5 years

@NeurIPSConf will run our workshop on Graph Representation Learning - details to follow @williamleif @thomaskipf @PetarV_93 @jure #geometricdeeplearning

0

7

51

2

17

168

@tkipf

Thomas Kipf

4 years

GraphMask looks like a very elegant approach to generate post-hoc explanations for Graph Neural Networks

@nicola_decao

Nicola De Cao

4 years

New hot pre-print 🔥Interpreting Graph Neural Networks for NLP With Differentiable Edge Masking🔥 We show you can learn to remove most of the edges in GNNs such that the remaning ones are interpretable! with @michael_sejr @iatitov

Tweet media one

Tweet media two

Tweet media three

Tweet media four

4

71

226

1

50

169

@tkipf

Thomas Kipf

2 months

Legends

Tweet media one

1

3

170

@tkipf

Thomas Kipf

2 years

Looking forward to presenting our work on Slot Attention for Video (SAVi) at @iclr_conf ! Slot-based neural architectures (think of them as latent GNNs) are one the latest developments in ML I am most excited about. Lots of great submissions in this area at ICLR this year!

@tkipf

Thomas Kipf

3 years

Excited to share our work on Conditional Object-Centric Learning from Video! We introduce SAVi, a slot-based model that can discover + represent visual entities in videos, using simple location cues and object motion (...or entirely unsupervised) 🖥️ 1/7

5

61

360

4

20

165

@tkipf

Thomas Kipf

2 years

Our latest work at @GoogleAI on emergent object-centric learning will be presented at @NeurIPSConf this year: 1) OSRT: Object Scene Representation Transformer

@tkipf

Thomas Kipf

2 years

So excited to share Object Scene Representation Transformer (OSRT): OSRT learns about complex 3D scenes & decomposes them into objects w/o supervision, while rendering novel views up to 3000x faster than prior methods! 🖥️ 📜 1/7

5

117

646

1

20

162

@tkipf

Thomas Kipf

3 years

An exciting new book draft that presents a unified view of Deep Learning architectures from a viewpoint of geometry and symmetry.

@PetarV_93

Petar Veličković

3 years

Proud to share our 150-page "proto-book" with @mmbronstein @joanbruna @TacoCohen on geometric DL! Through the lens of symmetries and invariances, we attempt to distill "all you need to build the architectures that are all you need". More info below! 🧵

Tweet media one

26

534

2K

0

24

162

@tkipf

Thomas Kipf

4 years

Want to learn more about object discovery, contrastive learning and world models? Visit our virtual poster at #ICLR2020 (Thurs, Sessions 1&2) on Contrastive Learning with Structured World Models, w/ @ElisevanderPol

Tweet media one

1

25

159

@tkipf

Thomas Kipf

5 years

The camera-ready version of our CompILE @icmlconf paper is out! Differentiable sequence segmentation for option discovery in RL — w/ @liyuajia @egrefen @pushmeet @PeterWBattaglia et al. Paper: Code:

Tweet media one

1

35

156

@tkipf

Thomas Kipf

6 years

Graph Convolutional Nets with Multi-Head Attention for Multi-Agent RL:

Tweet media one

0

38

152

@tkipf

Thomas Kipf

2 years

The call for papers for the workshop on Neuro Causal & Symbolic AI (nCSI) at @NeurIPSConf is now live! Submission deadline: 22 September 🖥️

Tweet media one

1

31

149

@tkipf

Thomas Kipf

1 year

Happy to share our #ICML2023 paper on Invariant Slot Attention (ISA)! ISA learns per-slot reference frames, enabling pose control of objects & more data efficient learning. 📜 💻 🖼️ 1/2

1

26

147

@tkipf

Thomas Kipf

7 years

Attentive graph neural nets are one of my favorite developments recently. Some noteworthy papers along these lines... 1/5

6

24

146

@tkipf

Thomas Kipf

4 years

Most comfortable conference experience so far! Watching @iclr_conf #africaNLP with @maartjeterhoeve (we both have a conference registration 🙃) from the comfort of our living room. Huge thanks to the conference/workshop organizers for making this happen!

Tweet media one

0

5

146

@tkipf

Thomas Kipf

1 year

Excited to share SlotFormer! SlotFormer provides a simple recipe for visual simulation from pixels: 1) Learn a slot-based representation of a video 2) Simulate slot dynamics using a Transformer (Led by @Dazitu_616 , to be presented at ICLR 2023) 1/6

1

29

143

@tkipf

Thomas Kipf

7 years

Graph Convolutional Networks v2: our latest work on modeling directed, relational graphs with GCNs

Tweet media one

1

50

141

@tkipf

Thomas Kipf

7 months

Happy to announce a Student Researcher opportunity in our team @GoogleDeepMind in Mountain View, California ☀️ If you’re excited about foundational research on controllability of visual/3D generative models, please consider applying:

Tweet card media

At the heart of Google DeepMind’s mission is our commitment to act as responsible pioneers in the field of AI, in service of society’s needs and expectations.

deepmind.google

3

21

144

@tkipf

Thomas Kipf

6 years

CompILE discovers composable segments & encodings of behavior from sequential data (unsupervised and differentiable). Codes can be recomposed to facilitate generalization and exploration in RL. New work with collaborators from @DeepMindAI Paper:

Tweet media one

2

31

143

@tkipf

Thomas Kipf

5 years

An extended version of our CompILE paper was accepted to @icmlconf ! See below for our earlier NeurIPS workshop version of the paper.

@tkipf

Thomas Kipf

6 years

CompILE discovers composable segments & encodings of behavior from sequential data (unsupervised and differentiable). Codes can be recomposed to facilitate generalization and exploration in RL. New work with collaborators from @DeepMindAI Paper:

Tweet media one

2

31

143

0

22

140

@tkipf

Thomas Kipf

4 years

I will be talking about Relational Structure Discovery today at the Graph Representation Learning (and Beyond) workshop at ICML. The talk will be live-streamed at 9:30 CEST. Join us (w/ @xbresson ) for live Q&A after the talk! Slide credit: @PetarV_93

Tweet media one

1

12

137

@tkipf

Thomas Kipf

6 years

Lesson learned: don’t give up on a rejected NIPS submission

@pbloemesquire

Peter Bloem (@[email protected])

6 years

Our R-GCN paper won the best student paper award at #ESWC ! Congrats to the best students Michael Schlichtkrull and @thomaskipf .

Tweet media one

1

19

71

2

30

140

@tkipf

Thomas Kipf

4 years

All you need is a graph neural net (for learning generalizable policies for structured agents). This is an interesting & exciting iteration on NerveNet () and its follow-up works.

@pathak2206

Deepak Pathak

4 years

RL gets specific to the robot it is trained on. Can a policy be trained to control many agents? Turns out, training (shared) policy for each motor instead of whole robot not only achieves SOTA at train but also transfers to unseen agents w/o fine-tuning!

7

237

965

3

26

141

@tkipf

Thomas Kipf

4 years

Accepted papers at the ELLIS Workshop on Geometric and Relational Deep Learning are now available on the workshop website: (incl. spotlight video and code for some)

Tweet media one

@tkipf

Thomas Kipf

4 years

We are organizing an @ELLISforEurope Workshop on Geometric and Relational Deep Learning! Registration invites will be shared soon. Interested in participating? Consider submitting an abstract or get in touch: w/ @erikjbekkers @wellingmax @mmbronstein

Tweet media one

3

81

279

2

35

134

@tkipf

Thomas Kipf

2 months

ICLR ( @iclr_conf ) this year had the best poster printing experience of any conference I’ve been to.

6

8

130

@tkipf

Thomas Kipf

2 months

🔥

Tweet media one

@gklambauer

Günter Klambauer

2 months

xLSTM: Extended Long Short-Term Memory The famous LSTM nets are improved by exponential gates and matrix-memory with covariance update. Strong results on large-scale language modeling. P:

Tweet media one

5

49

211

1

16

126

@tkipf

Thomas Kipf

7 years

New, much improved version of our Graph Convolutional Matrix Completion paper is now on arXiv (with @vdbergrianne )

Tweet media one

1

35

125

@tkipf

Thomas Kipf

1 year

Excited to announce RUST (Really Unposed SRT): RUST learns generalizable scene representations from raw images: no camera poses needed! A latent pose allows us to control the camera at test time 🎥 🖥️ 📜 CVPR'23 (Highlight) 1/7

3

25

124

@tkipf

Thomas Kipf

2 months

Special thanks to @jo_brandstetter for showing us his home turf after an amazing week at @iclr_conf in Vienna. 🇦🇹🏔️

Tweet media one

Tweet media two

Tweet media three

Tweet media four

9

1

123

@tkipf

Thomas Kipf

9 months

Couldn't agree more: we still don't have good solutions for scalable abstraction/concept learning from "raw" data (be it language, vision, or other sensory data). Solving this would likely unlock important new capabilities.

@iamtrask

Andrew Trask

9 months

For anyone interested in future LLM development One of the bigger unsolved deep learning problems: learning of hierarchical structure Example: we still use tokenizers to train SOTA LLMs. We should be able to feed in bits/chars/bytes and get SOTA Related: larger context window

19

76

520

4

16

120

@tkipf

Thomas Kipf

6 years

This development is highly concerning, and makes it extremely difficult for candidates from non-CS backgrounds to get into these programs. I myself was lucky enough to get into a ML PhD program two years ago although my background was in Physics. Not sure if still possible today.

2

22

117

@tkipf

Thomas Kipf

1 year

(Emergent) semantic segmentation from image-text pairs using Slot Attention Jilan Xu, @WeidiXie et al.: Learning Open-vocabulary Semantic Segmentation Models From Natural Language Supervision 🖥️ 📜

Tweet media one

3

23

115