Workshop
Differentiable Almost Everything: Differentiable Relaxations, Algorithms, Operators, and Simulators
Felix Petersen 路 Marco Cuturi 路 Hilde Kuehne 路 Christian Borgelt 路 Lawrence Stewart 路 Michael Kagan 路 Stefano Ermon
Stolz 0
Fri 26 Jul, midnight PDT
Gradients and derivatives are integral to machine learning, as they enable gradient-based optimization. In many real applications, however, models rest on algorithmic components that implement discrete decisions, or rely on discrete intermediate representations and structures. These discrete steps are intrinsically non-differentiable and accordingly break the flow of gradients. To use gradient-based approaches to learn the parameters of such models requires turning these non-differentiable components differentiable. This can be done with careful considerations, notably, using smoothing or relaxations to propose differentiable proxies for these components. With the advent of modular deep learning frameworks, these ideas have become more popular than ever in many fields of machine learning, generating in a short time-span a multitude of "differentiable everything", impacting topics as varied as rendering, sorting and ranking, convex optimizers, shortest-paths, dynamic programming, physics simulations, NN architecture search, top-k, graph algorithms, weakly- and self-supervised learning, and many more.
Schedule
Fri 12:00 a.m. - 12:10 a.m.
|
Opening Remarks
SlidesLive Video |
Felix Petersen 馃敆 |
Fri 12:10 a.m. - 12:50 a.m.
|
Invited Talk 1: The Challenges of Differentiating through Logic
(
Invited Talk
)
>
SlidesLive Video |
Emile van Krieken 馃敆 |
Fri 12:50 a.m. - 1:25 a.m.
|
Invited Talk 2: Diffusion Models powered by Differentiable Simulations
(
Invited Talk
)
>
SlidesLive Video |
Nils Thuerey 馃敆 |
Fri 1:25 a.m. - 1:45 a.m.
|
Coffee Break
|
馃敆 |
Fri 1:45 a.m. - 2:10 a.m.
|
Invited Talk 3: Differentiable Control Flows and Data Structures
(
Invited Talk
)
>
SlidesLive Video |
Vincent Roulet 馃敆 |
Fri 2:10 a.m. - 2:45 a.m.
|
Short Poster Talks 1
(
Short Poster Talks
)
>
SlidesLive Video |
馃敆 |
Fri 2:45 a.m. - 3:45 a.m.
|
Poster Session 1
(
Poster Session
)
>
|
馃敆 |
Fri 3:45 a.m. - 4:40 a.m.
|
Lunch Break
|
馃敆 |
Fri 4:40 a.m. - 5:35 a.m.
|
Invited Talk 4: Is Differentiating Almost Everything Really the Only Solution?
SlidesLive Video |
Francis Bach 馃敆 |
Fri 5:35 a.m. - 5:55 a.m.
|
Coffee Break
|
馃敆 |
Fri 5:55 a.m. - 6:10 a.m.
|
Short Poster Talks 2
(
Short Poster Talks
)
>
SlidesLive Video |
馃敆 |
Fri 6:10 a.m. - 6:50 a.m.
|
Invited Talk 5: Progress and Prospects in Differentiable Simulation for Robotics
(
Invited Talk
)
>
SlidesLive Video |
Justin Carpentier 馃敆 |
Fri 6:50 a.m. - 7:00 a.m.
|
Closing Remarks
SlidesLive Video |
Felix Petersen 馃敆 |
Fri 7:00 a.m. - 8:00 a.m.
|
Poster Session 2
(
Poster Session
)
>
|
馃敆 |
-
|
Differentiable Soft Min-Max Loss to Restrict Weight Range for Model Quantization ( Poster ) > link | Arnav Kundu 路 Chungkuk Yoo 路 Minsik Cho 路 Saurabh Adya 馃敆 |
-
|
Learning to Design Data-structures: A Case Study of Nearest Neighbor Search ( Poster ) > link | Omar Salemohamed 路 Vatsal Sharan 路 Shivam Garg 路 Laurent Charlin 路 Greg Valiant 馃敆 |
-
|
Transforming a Non-Differentiable Rasterizer into a Differentiable One with Stochastic Gradient Estimation ( Poster ) > link | Thomas Deliot 路 Eric Heitz 路 Laurent Belcour 馃敆 |
-
|
Relaxing Graph Transformers for Adversarial Attacks ( Poster ) > link | Philipp Foth 路 Lukas Gosch 路 Simon Markus Geisler 路 Leo Schwinn 路 Stephan G眉nnemann 馃敆 |
-
|
You Shall Pass: Dealing with the Zero-Gradient Problem in Predict and Optimize for Convex Optimization ( Poster ) > link | Grigorii Veviurko 路 Wendelin Boehmer 路 Mathijs de Weerdt 馃敆 |
-
|
PICT: Adaptive GPU Accelerated Differentiable Fluid Simulation for Machine Learning ( Poster ) > link | Erik Franz 路 Nils Thuerey 馃敆 |
-
|
Revisiting Score Function Estimators for kk-Subset Sampling ( Poster ) > link | Klas Wijk 路 Ricardo Vinuesa 路 Hossein Azizpour 馃敆 |
-
|
Differentiable Mapper for Topological Optimization of Data Representation ( Poster ) > link | Ziyad Oulhaj 路 Mathieu Carri猫re 路 Bertrand Michel 馃敆 |
-
|
Differentiable Iterated Function Systems ( Poster ) > link | Cory Scott 馃敆 |
-
|
Differentiable Weighted Automata ( Poster ) > link | Anand Balakrishnan 路 Jyotirmoy Deshmukh 馃敆 |
-
|
Learning Set Functions with Implicit Differentiation ( Poster ) > link | G枚zde 脰zcan 路 Chengzhi Shi 路 Stratis Ioannidis 馃敆 |
-
|
SA-DQAS: Self-attention Enhanced Differentiable Quantum Architecture Search ( Poster ) > link | Yize Sun 路 Jiarui Liu 路 Zixin Wu 路 Zifeng Ding 路 Yunpu Ma 路 Thomas Seidl 路 Volker Tresp 馃敆 |
-
|
Differentiable Local Intrinsic Dimension Estimation with Diffusion Models ( Poster ) > link | Hamidreza Kamkari 路 Brendan Ross 路 Rasa Hosseinzadeh 路 Jesse Cresswell 路 Gabriel Loaiza-Ganem 馃敆 |
-
|
Using gradients to check sensitivity of MCMC-based analyses to removing data ( Poster ) > link | Tin Nguyen 路 Ryan Giordano 路 Rachael Meager 路 Tamara Broderick 馃敆 |
-
|
A Differentiable Approach to Multi-scale Brain Modeling ( Poster ) > link | Chaoming Wang 路 Muyang Lyu 路 Tianqiu Zhang 路 Sichao He 路 Si Wu 馃敆 |
-
|
End-to-end Differentiable Model of Robot-terrain Interactions ( Poster ) > link | Ruslan Agishev 路 Vladim铆r Kubelka 路 Martin Pecka 路 Tomas Svoboda 路 Karel Zimmermann 馃敆 |
-
|
Differentiable Cost-Parameterized Monge Map Estimators ( Poster ) > link | Samuel Howard 路 George Deligiannidis 路 Patrick Rebeschini 路 James Thornton 馃敆 |
-
|
Symbolic Autoencoding for Self-Supervised Sequence Learning ( Poster ) > link | Mohammad Hossein Amani 路 Nicolas Baldwin 路 Amin Mansouri 路 Martin Josifoski 路 Maxime Peyrard 路 Robert West 馃敆 |
-
|
Generalizing Convolution to Point Clouds ( Poster ) > link | Davide Bacciu 路 Francesco Landolfi 馃敆 |
-
|
Differentiable Cluster Graph Neural Network ( Poster ) > link | Yanfei Dong 路 Mohammed Haroon Dupty 路 Lambert Deng 路 Zhuanghua Liu 路 Yong Liang Goh 路 Wee Sun Lee 馃敆 |
-
|
CGMTorch: A Framework for Gradient-based Design of Computational Granular Metamaterials ( Poster ) > link | Atoosa Parsa 路 Corey OHern 路 Rebecca Kramer-Bottiglio 路 Josh Bongard 馃敆 |
-
|
(Almost) Smooth Sailing: Steering Towards Differentiable Regularization for Stability ( Poster ) > link | Rossen Nenov 路 Daniel Haider 路 Peter Balazs 馃敆 |
-
|
Differentiable Approximations of Fair OWA Optimization ( Poster ) > link | My Dinh 路 James Kotary 路 Ferdinando Fioretto 馃敆 |
-
|
Implicit Diffusion: Efficient Optimization through Stochastic Sampling ( Poster ) > link | Pierre Marion 路 Anna Korba 路 Peter Bartlett 路 Mathieu Blondel 路 Valentin De Bortoli 路 Arnaud Doucet 路 Felipe Llinares-Lopez 路 Courtney Paquette 路 Quentin Berthet 馃敆 |
-
|
BMapEst: Estimation of Brain Tissue Probability Maps using a Differentiable MRI Simulator ( Poster ) > link | Utkarsh Gupta 路 Emmanouil Nikolakakis 路 Moritz Zaiss 路 Razvan Marinescu 馃敆 |
-
|
Stable Differentiable Causal Discovery ( Poster ) > link | Achille Nazaret 路 Justin Hong 路 Elham Azizi 路 David Blei 馃敆 |
-
|
Differentiable Wireless Simulation with Geometric Transformers ( Poster ) > link | Thomas Hehn 路 Markus Peschl 路 Tribhuvanesh Orekondy 路 Arash Behboodi 路 Johann Brehmer 馃敆 |
-
|
MAGNOLIA: Matching Algorithms via GNNs for Online Value-to-go Approximation ( Poster ) > link | Alexandre Hayderi 路 Amin Saberi 路 Ellen Vitercik 路 Anders Wikum 馃敆 |
-
|
Enhancing Concept-based Learning with Logic ( Poster ) > link | Deepika Vemuri 路 Gautham Bellamkonda 路 Vineeth N Balasubramanian 馃敆 |
-
|
Graph Neural Networks for Binary Programming ( Poster ) > link | Moshe Eliasof 路 Eldad Haber 馃敆 |
-
|
How Consensus-Based Optimization can be Interpreted as a Stochastic Relaxation of Gradient Descent ( Poster ) > link | Konstantin Riedl 路 Timo Klock 路 Carina Geldhauser 路 Massimo Fornasier 馃敆 |
-
|
DiffFit: Differentiable Fitting of Molecule Structures to Cryo-EM Map ( Poster ) > link | Deng Luo 路 Zainab Alsuwaykit 路 Dawar Khan 路 Ondrej Strnad 路 Tobias Isenberg 路 Ivan Viola 馃敆 |
-
|
桅Flow桅Flow: Differentiable Simulations for Machine Learning ( Poster ) > link | Philipp Holl 路 Nils Thuerey 馃敆 |
-
|
Heterogeneous Federated Zeroth-Order Optimization using Gradient Surrogates ( Poster ) > link | Yao Shu 路 Xiaoqiang Lin 路 Zhongxiang Dai 路 Bryan Kian Hsiang Low 馃敆 |
-
|
Decoupled Differentiable Neural Architecture Search: Memory-Efficient Differentiable NAS via Disentangled Search Space ( Poster ) > link | Libin Hou 馃敆 |
-
|
A framework for differentiable Supervised Graph Prediction ( Poster ) > link | Paul KRZAKALA 路 Junjie Yang 路 R茅mi Flamary 路 Florence d'Alch茅-Buc 路 Charlotte Laclau 路 Matthieu Labeau 馃敆 |
-
|
A Differentiable Topological Notion of Local Maxima for Keypoint Detection ( Poster ) > link | Giovanni Barbarani 路 Francesco Vaccarino 路 Gabriele Trivigno 路 Marco Guerra 路 Gabriele Berton 路 Carlo Masone 馃敆 |
-
|
Parallelising Differentiable Algorithms Removes the Scalar Bottleneck: A Case Study ( Poster ) > link | Euan Ong 路 Ferenc Husz谩r 路 Petar Veli膷kovi膰 馃敆 |
-
|
BPNAS: Bayesian Progressive Neural Architecture Search ( Poster ) > link | Hyunwoong Chang 路 Anirban Samaddar 路 Sandeep Madireddy 馃敆 |
-
|
Analyzing and Improving Surrogate Gradient Training in Discrete Neural Networks Using Dynamical Systems Theory ( Poster ) > link | Rainer Engelken 路 Larry Abbott 馃敆 |
-
|
Better Structure- and Function-Aware Substitution Matrices via Differentiable Graph Matching ( Poster ) > link | Paolo Pellizzoni 路 Carlos Oliver 路 Karsten Borgwardt 馃敆 |
-
|
Differentiable Short-Time Fourier Transform: A Time-Frequency Layer with Learnable Parameters ( Poster ) > link | Maxime Leiber 路 Yosra MARNISSI 路 Axel Barrau 馃敆 |
-
|
Energy-based Hopfield Boosting for Out-of-Distribution Detection ( Poster ) > link | Claus Hofmann 路 Simon Schmid 路 Bernhard Lehner 路 Daniel Klotz 路 Sepp Hochreiter 馃敆 |
-
|
BiPer: Binary Neural Networks using a Periodic Function ( Poster ) > link | Edwin Vargas 路 Claudia Correa 路 Carlos Hinojosa 路 Henry Arguello 馃敆 |