Skip to content

Commit c6897dd

Browse files
o-laurento-laurent
committed
✨ Add schedule
1 parent 0e60a60 commit c6897dd

File tree

1 file changed

+155
-129
lines changed

1 file changed

+155
-129
lines changed

‎index.html

Lines changed: 155 additions & 129 deletions
Original file line numberDiff line numberDiff line change
@@ -32,8 +32,8 @@
3232
<h1 class="project-name">A Bayesian Odyssey in Uncertainty: from Theoretical
3333
Foundations to Real-World Applications</h1>
3434
<h2 class="project-tagline">ECCV 2024 - Room: <strong>Suite 7</strong><br><strong>30 Sept</strong> - 8:30 AM to
35-
12:30 PM<br>This tutorial will be
36-
<strong>available online and recorded</strong>
35+
12:30 PM<br>This recording of the tutorial will be
36+
<strong>available online</strong>.
3737
</h2>
3838
</section>
3939

@@ -97,146 +97,172 @@ <h2 style="text-align: center">Overview</h2>
9797
</p>
9898
</div>
9999

100-
<br>
101-
102100
<div class="containertext" style="max-width:50rem">
103-
<h2 style="text-align: center">Outline</h2>
104-
105-
<h3 style="text-align: left">Introduction: Why & where is UQ helpful?</h3>
106-
<p>
107-
Initial exploration into the critical role of uncertainty quantification (UQ) within the realm
108-
of computer vision (CV): participants will gain an understanding of why it’s essential to consider
109-
uncertainty in CV, especially concerning decision-making in complex
110-
environments. We will introduce real-world scenarios where uncertainty can profoundly
111-
impact model performance and safety, setting the stage for deeper exploration through out the tutorial.
112-
</p>
113-
<h3 style="text-align: left">From maximum a posteriori to BNNs.</h3>
114-
<p>
115-
In this part, we will journey through the evolution of UQ techniques, starting
116-
from classic approaches such as maximum a posteriori estimation to the more ellaborate Bayesian Neural
117-
Networks. The participants will grasp the conceptual foundations
118-
of UQ, laying the groundwork for the subsequent discussions of Bayesian methods.
119-
</p>
120-
<h3 style="text-align: left">Strategies for BNN posterior inference.</h3>
121-
<p>
122-
This is the core part, which will dive into the process of estimating the posterior distribution of BNNs.
123-
The participants
124-
will gain insights into the computational complexities involved in modeling uncertainty
125-
through a comprehensive overview of techniques such as Variational Inference (VI),
126-
Hamiltonian Monte Carlo (HMC), and Langevin Dynamics. Moreover, we will explore
127-
the characteristics and visual representation of posterior distributions, providing a better
128-
understanding of Bayesian inference.
129-
</p>
130-
<h3 style="text-align: left">Computationally-efficient BNNs for CV.</h3>
101+
<h2 style="text-align: center">Schedule</h2>
131102
<p>
132-
Here, we will present recent techniques to improve the computational efficiency of BNNs for computer vision
133-
tasks.
134-
We will present different forms of obtaining BNNs from a intermediate checkpoints,
135-
weight trajectories during a training run, different types of variational subnetworks,
136-
etc., along with their main strenghts and limitations.
137-
</p>
138-
<h3 style="text-align: left">Convert your DNN into a BNN: post-hoc BNN inference.</h3>
139-
<p>
140-
This segment focuses on post-hoc inference techniques, with a focus on Laplace approximation. The
141-
participants
142-
will learn how Laplace approximation serves as a computationally efficient method for
143-
approximating the posterior distribution of Bayesian Neural Networks.
144-
</p>
145-
<h3 style="text-align: left">Quality of estimated uncertainty and practical examples.</h3>
146-
<p>
147-
In the final session, participants will learn how to evaluate the quality of UQ in practi-
148-
cal settings. We will develop multiple approaches to assess the reliability and calibra-
149-
tion of uncertainty estimates, equipping participants with the tools to gauge the robust-
150-
ness of their models. Additionally, we will dive into real-world examples and applica-
151-
tions, showcasing how UQ can enhance the reliability
152-
and performance of computer vision systems in diverse scenarios. Through interactive
153-
discussions and case studies, participants will gain practical insights into deploying
154-
uncertainty-aware models in real-world applications.
155-
</p>
103+
<ul>
104+
<li>8:45-9:15: Opening - Andrei</li>
105+
<li>
106+
9:15-10:05: Uncertainty quantification: from maximum a posteriori to BNNs - Pavel (remotely)
107+
</li>
108+
<li>10:05-10:30: Computationally-efficient BNNs for computer vision - Gianni</li>
109+
<li>10:35-11:00: Coffee</li>
110+
<li>11:00-11:50: Convert your DNN into a BNN - Alexander</li>
111+
<li>11:50-12:20: Quality of estimated uncertainty and practical examples - Adrien (remotely) & Gianni </li>
112+
<li>12:20-12:40: Closing remarks + Q&A - Andrei, Alex, Pavel & Gianni</li>
113+
</ul>
156114

157-
<h3 style="text-align: left">Uncertainty Quantification Framework.</h3>
158-
<p>
159-
This tutorial will also very quickly introduce the <a
160-
href="https://github.com/ensta-u2is-ai/torch-uncertainty">TorchUncertainty
161-
library</a>, an uncertainty-aware open-source framework for training models in PyTorch.
162-
</p>
163-
</div>
164115

165-
<a href="https://torch-uncertainty.github.io/" target="_blank">
166-
<div><img src="assets/logoTU_full.png" width="20%" hspace="2%"> </div>
167-
</a>
168116

169-
<br>
170117

171-
<div class="containertext" style="max-width:50rem">
172-
<h2 style="text-align: center">Relation to prior tutorials and short courses</h2>
173-
<p> This tutorial is affiliated with the <a href="https://uncv2023.github.io/">UNCV Workshop</a>,
174-
which had its inaugural edition at ECCV 2022, a subsequent one at ICCV, and is back at ECCV this year.
175-
In constrast to the workshop, the tutorial puts its primary emphasis on the theoretical facets. </p>
176-
<p> UQ has received some attention
177-
in recent times, as evidenced by its inclusion in
178-
the tutorial <a href="https://abursuc.github.io/many-faces-reliability/">'Many Faces of Reliability of Deep
179-
Learning for Real-World Deployment'</a>. While this tutorial explored various applications associated with
180-
uncertainty,
181-
it did not place a specific emphasis on probabilistic models and Bayesian Neural Networks. Our tutorial aims
182-
to provide a more in-depth exploration of uncertainty theory, accompanied by the introduction of practical
183-
applications, including the presentation of the library, <a
184-
href="https://github.com/ensta-u2is-ai/torch-uncertainty">TorchUncertainty</a>.</p>
185-
</div>
186118

187-
<div class="containertext" style="max-width:50rem">
188-
<h2 style="text-align: center">Selected References</h2>
189-
<ol>
190-
<li><b>Immer, A.</b>, Palumbo, E., Marx, A., & Vogt, J. E. E<a
191-
href="https://proceedings.neurips.cc/paper_files/paper/2023/file/a901d5540789a086ee0881a82211b63d-Paper-Conference.pdf">
192-
Effective Bayesian Heteroscedastic Regres-
193-
sion with Deep Neural Networks</a>. In NeurIPS, 2023.</li>
194-
<li><b>Franchi, G., Bursuc, A.,</b> Aldea, E., Dubuisson, S.,
195-
& Bloch, I. <a href="https://arxiv.org/pdf/2012.02818">Encoding the latent posterior of
196-
Bayesian Neural Networks for uncertainty quantification</a>. IEEE TPAMI, 2023.</li>
197-
<li><b>Franchi, G.</b>, Yu, X., <b>Bursuc, A.</b>, Aldea, E., Dubuisson,
198-
S., & Filliat, D. <a href="https://arxiv.org/pdf/2207.10130">Latent Discriminant
199-
deterministic Uncertainty</a>. In ECCV 2022.</li>
200-
<li><b>Laurent, O.</b>, <b>Lafage, A.</b>, Tartaglione, E., Daniel, G.,
201-
Martinez, J. M., <b>Bursuc, A.</b>, & <b>Franchi, G.</b>
202-
<a href="https://arxiv.org/pdf/2210.09184">Packed-Ensembles for Efficient Uncertainty Estimation</a>. In
203-
ICLR 2023.
204-
</li>
205-
<li><b>Izmailov, P.</b>, Vikram, S., Hoffman, M. D., & Wilson, A. G. <a
206-
href="https://arxiv.org/pdf/2104.14421">What are Bayesian neural network
207-
posteriors really like?</a> In ICML, 2021.</li>
208-
<li><b>Izmailov, P.</b>, Maddox, W. J., Kirichenko, P., Garipov, T., Vetrov, D., & Wilson, A. G. <a
209-
href="https://arxiv.org/pdf/1907.07504">Subspace inference for Bayesian deep learning</a>. In UAI, 2020.
210-
</li>
211-
<li><b>Franchi, G.</b>, <b>Bursuc, A.</b>, Aldea, E., Dubuisson, S., &
212-
Bloch, I. <a href="https://arxiv.org/pdf/1912.11316">TRADI: Tracking deep neural
213-
network weight distributions</a>. In ECCV 2020.</li>
214-
<li>Wilson, A. G., & <b>Izmailov, P</b>. <a href="https://arxiv.org/pdf/2002.08791">Bayesian deep
215-
learning and a probabilistic perspective of generalization</a>. In NeurIPS, 2020.</li>
216-
<li>Hendrycks, D., Dietterich, T. <a href="https://arxiv.org/pdf/1903.12261">Benchmarking Neural Network
217-
Robustness to Common Corruptions and
218-
Perturbations</a>. In ICLR 2019.</li>
219-
<li><b> Izmailov, P.</b>, Podoprikhin, D., Garipov, T., Vetrov, D., & Wilson, A. G. <a
220-
href="https://arxiv.org/pdf/1803.05407">Averaging weights
221-
leads to wider optima and better generalization</a>. In UAI, 2018. </li>
222-
</ol>
223-
You will find more references in the <a
224-
href="https://github.com/ensta-u2is-ai/awesome-uncertainty-deeplearning">Awesome Uncertainty in deep
225-
learning.</a>
226-
</div>
227119

228-
<br>
120+
</p>
121+
<br>
229122

230-
<div class="containertext">
231-
<h3 style="text-align: center">Andrei Bursuc is supported by ELSA:</h3>
123+
<div class="containertext" style="max-width:50rem">
124+
<h2 style="text-align: center">Outline</h2>
232125

233-
<center>
234-
<a href="https://elsa-ai.eu/" target="_blank"><img src="assets/elsa_logo.png" width="10%" hspace="2%" />
235-
</center>
126+
<h3 style="text-align: left">Introduction: Why & where is UQ helpful?</h3>
127+
<p>
128+
Initial exploration into the critical role of uncertainty quantification (UQ) within the realm
129+
of computer vision (CV): participants will gain an understanding of why it’s essential to consider
130+
uncertainty in CV, especially concerning decision-making in complex
131+
environments. We will introduce real-world scenarios where uncertainty can profoundly
132+
impact model performance and safety, setting the stage for deeper exploration through out the tutorial.
133+
</p>
134+
<h3 style="text-align: left">From maximum a posteriori to BNNs.</h3>
135+
<p>
136+
In this part, we will journey through the evolution of UQ techniques, starting
137+
from classic approaches such as maximum a posteriori estimation to the more ellaborate Bayesian Neural
138+
Networks. The participants will grasp the conceptual foundations
139+
of UQ, laying the groundwork for the subsequent discussions of Bayesian methods.
140+
</p>
141+
<h3 style="text-align: left">Strategies for BNN posterior inference.</h3>
142+
<p>
143+
This is the core part, which will dive into the process of estimating the posterior distribution of BNNs.
144+
The participants
145+
will gain insights into the computational complexities involved in modeling uncertainty
146+
through a comprehensive overview of techniques such as Variational Inference (VI),
147+
Hamiltonian Monte Carlo (HMC), and Langevin Dynamics. Moreover, we will explore
148+
the characteristics and visual representation of posterior distributions, providing a better
149+
understanding of Bayesian inference.
150+
</p>
151+
<h3 style="text-align: left">Computationally-efficient BNNs for CV.</h3>
152+
<p>
153+
Here, we will present recent techniques to improve the computational efficiency of BNNs for computer
154+
vision
155+
tasks.
156+
We will present different forms of obtaining BNNs from a intermediate checkpoints,
157+
weight trajectories during a training run, different types of variational subnetworks,
158+
etc., along with their main strenghts and limitations.
159+
</p>
160+
<h3 style="text-align: left">Convert your DNN into a BNN: post-hoc BNN inference.</h3>
161+
<p>
162+
This segment focuses on post-hoc inference techniques, with a focus on Laplace approximation. The
163+
participants
164+
will learn how Laplace approximation serves as a computationally efficient method for
165+
approximating the posterior distribution of Bayesian Neural Networks.
166+
</p>
167+
<h3 style="text-align: left">Quality of estimated uncertainty and practical examples.</h3>
168+
<p>
169+
In the final session, participants will learn how to evaluate the quality of UQ in practi-
170+
cal settings. We will develop multiple approaches to assess the reliability and calibra-
171+
tion of uncertainty estimates, equipping participants with the tools to gauge the robust-
172+
ness of their models. Additionally, we will dive into real-world examples and applica-
173+
tions, showcasing how UQ can enhance the reliability
174+
and performance of computer vision systems in diverse scenarios. Through interactive
175+
discussions and case studies, participants will gain practical insights into deploying
176+
uncertainty-aware models in real-world applications.
177+
</p>
178+
179+
<h3 style="text-align: left">Uncertainty Quantification Framework.</h3>
180+
<p>
181+
This tutorial will also very quickly introduce the <a
182+
href="https://github.com/ensta-u2is-ai/torch-uncertainty">TorchUncertainty
183+
library</a>, an uncertainty-aware open-source framework for training models in PyTorch.
184+
</p>
185+
</div>
186+
187+
<a href="https://torch-uncertainty.github.io/" target="_blank">
188+
<div><img src="assets/logoTU_full.png" width="20%" hspace="2%"> </div>
236189
</a>
190+
191+
<br>
192+
193+
<div class="containertext" style="max-width:50rem">
194+
<h2 style="text-align: center">Relation to prior tutorials and short courses</h2>
195+
<p> This tutorial is affiliated with the <a href="https://uncv2023.github.io/">UNCV Workshop</a>,
196+
which had its inaugural edition at ECCV 2022, a subsequent one at ICCV, and is back at ECCV this year.
197+
In constrast to the workshop, the tutorial puts its primary emphasis on the theoretical facets. </p>
198+
<p> UQ has received some attention
199+
in recent times, as evidenced by its inclusion in
200+
the tutorial <a href="https://abursuc.github.io/many-faces-reliability/">'Many Faces of Reliability of
201+
Deep
202+
Learning for Real-World Deployment'</a>. While this tutorial explored various applications associated
203+
with
204+
uncertainty,
205+
it did not place a specific emphasis on probabilistic models and Bayesian Neural Networks. Our tutorial
206+
aims
207+
to provide a more in-depth exploration of uncertainty theory, accompanied by the introduction of practical
208+
applications, including the presentation of the library, <a
209+
href="https://github.com/ensta-u2is-ai/torch-uncertainty">TorchUncertainty</a>.</p>
210+
</div>
211+
212+
<div class="containertext" style="max-width:50rem">
213+
<h2 style="text-align: center">Selected References</h2>
214+
<ol>
215+
<li><b>Immer, A.</b>, Palumbo, E., Marx, A., & Vogt, J. E. E<a
216+
href="https://proceedings.neurips.cc/paper_files/paper/2023/file/a901d5540789a086ee0881a82211b63d-Paper-Conference.pdf">
217+
Effective Bayesian Heteroscedastic Regres-
218+
sion with Deep Neural Networks</a>. In NeurIPS, 2023.</li>
219+
<li><b>Franchi, G., Bursuc, A.,</b> Aldea, E., Dubuisson, S.,
220+
& Bloch, I. <a href="https://arxiv.org/pdf/2012.02818">Encoding the latent posterior of
221+
Bayesian Neural Networks for uncertainty quantification</a>. IEEE TPAMI, 2023.</li>
222+
<li><b>Franchi, G.</b>, Yu, X., <b>Bursuc, A.</b>, Aldea, E., Dubuisson,
223+
S., & Filliat, D. <a href="https://arxiv.org/pdf/2207.10130">Latent Discriminant
224+
deterministic Uncertainty</a>. In ECCV 2022.</li>
225+
<li><b>Laurent, O.</b>, <b>Lafage, A.</b>, Tartaglione, E., Daniel, G.,
226+
Martinez, J. M., <b>Bursuc, A.</b>, & <b>Franchi, G.</b>
227+
<a href="https://arxiv.org/pdf/2210.09184">Packed-Ensembles for Efficient Uncertainty Estimation</a>. In
228+
ICLR 2023.
229+
</li>
230+
<li><b>Izmailov, P.</b>, Vikram, S., Hoffman, M. D., & Wilson, A. G. <a
231+
href="https://arxiv.org/pdf/2104.14421">What are Bayesian neural network
232+
posteriors really like?</a> In ICML, 2021.</li>
233+
<li><b>Izmailov, P.</b>, Maddox, W. J., Kirichenko, P., Garipov, T., Vetrov, D., & Wilson, A. G. <a
234+
href="https://arxiv.org/pdf/1907.07504">Subspace inference for Bayesian deep learning</a>. In UAI,
235+
2020.
236+
</li>
237+
<li><b>Franchi, G.</b>, <b>Bursuc, A.</b>, Aldea, E., Dubuisson, S., &
238+
Bloch, I. <a href="https://arxiv.org/pdf/1912.11316">TRADI: Tracking deep neural
239+
network weight distributions</a>. In ECCV 2020.</li>
240+
<li>Wilson, A. G., & <b>Izmailov, P</b>. <a href="https://arxiv.org/pdf/2002.08791">Bayesian deep
241+
learning and a probabilistic perspective of generalization</a>. In NeurIPS, 2020.</li>
242+
<li>Hendrycks, D., Dietterich, T. <a href="https://arxiv.org/pdf/1903.12261">Benchmarking Neural Network
243+
Robustness to Common Corruptions and
244+
Perturbations</a>. In ICLR 2019.</li>
245+
<li><b> Izmailov, P.</b>, Podoprikhin, D., Garipov, T., Vetrov, D., & Wilson, A. G. <a
246+
href="https://arxiv.org/pdf/1803.05407">Averaging weights
247+
leads to wider optima and better generalization</a>. In UAI, 2018. </li>
248+
</ol>
249+
You will find more references in the <a
250+
href="https://github.com/ensta-u2is-ai/awesome-uncertainty-deeplearning">Awesome Uncertainty in deep
251+
learning.</a>
252+
</div>
253+
254+
<br>
255+
256+
<div class="containertext">
257+
<h3 style="text-align: center">Andrei Bursuc is supported by ELSA:</h3>
258+
259+
<center>
260+
<a href="https://elsa-ai.eu/" target="_blank"><img src="assets/elsa_logo.png" width="10%" hspace="2%" />
261+
</center>
262+
</a>
263+
</div>
237264
</div>
238265
</div>
239-
</div>
240266

241267

242268
</section>

0 commit comments

Comments
 (0)