Skip to content

Commit 545439a

Browse files
dzambondzambon
committed
create tutorial for the USI-SUPSI phd school 2026
1 parent 2468bf6 commit 545439a

1 file changed

Lines changed: 166 additions & 0 deletions

File tree

‎_tutorials/graph-based-processing/dsai-school-2026.html‎

Lines changed: 166 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,166 @@
1+
---
2+
layout: particles_header
3+
title: Graph Deep Learning for Time Series and Spatiotemporal Data
4+
description: The GMLG tutorial on graph deep learning for time-series processing.
5+
venue: Winter School in Network Data Science and Artificial Intelligence 2026
6+
city: Lugano, Switzerland
7+
---
8+
<!-- Outline section-->
9+
<section>
10+
<div class="container px-4">
11+
<div class="row gx-4 justify-content-center">
12+
<div class="col-lg-8">
13+
<p>
14+
Successful applications of <strong>deep learning</strong> in <strong>time series</strong> processing
15+
often involve training a single neural network on a collection of (related) time series.
16+
Pairwise relationships among time series can be modeled by considering a (possibly dynamic) graph
17+
spanning the collection. In this context, <strong>graph-based
18+
methods</strong> take the standard deep learning approach to time series processing a step
19+
forward.
20+
The recent theoretical and practical developments in graph machine learning make adopting such an
21+
approach particularly appealing and timely.
22+
23+
The twofold <strong>objective</strong> of this tutorial is to:
24+
<ol>
25+
<li>offer a <strong>comprehensive overview</strong> of the field, with a focus on forecasting
26+
applications;
27+
</li>
28+
<li>provide <strong>tools and guidelines</strong> to design and evaluate graph-based models for time
29+
series.
30+
</li>
31+
</ol>
32+
This tutorial is meant for early-career researchers and practitioners who wish to apply graph deep
33+
learning methods to their time series processing applications.
34+
At the same time, the tutorial provides experienced scholars with a coherent framing of the state of the art and new perspectives.
35+
</p>
36+
</div>
37+
<div class="col-lg-8 text-center">
38+
<h1>Material</h1>
39+
<p>Download the slides used in our tutorial.</p>
40+
<div class="row g-4 justify-content-center">
41+
<div class="col-md-4">
42+
<a class="w-75 btn btn-primary" href="https://drive.google.com/file/d/18tpZTHw-ZheWCeVLEgR9Nzo6USsY_O3_/view?usp=sharing" role="button" target="_blank">Download slides</a>
43+
</div>
44+
<div class="col-md-4">
45+
<a class="w-75 btn btn-warning" href="https://github.com/TorchSpatiotemporal/tsl/blob/main/examples/notebooks/a_gentle_introduction_to_tsl.ipynb" role="button" target="_blank">Code demo</a>
46+
</div>
47+
</div>
48+
</div>
49+
</div>
50+
</section>
51+
<!-- Location section-->
52+
<section class="text-white py-0"
53+
style="background-image: url(./img/.jpg); background-attachment: fixed; background-size: cover; background-position: center;">
54+
<div class="w-100 h-100 py-5" style="backdrop-filter: brightness(0.5) blur(3px);">
55+
<div class="container px-4">
56+
<div class="row gx-4 justify-content-center">
57+
<div class="col-lg-8">
58+
<p class="lead">
59+
This tutorial will be delivered at the <strong class="text-decoration-underline">Winter School in Network Data Science and Artificial Intelligence</strong>, held online <strong>from the 26th to the 29th of November
60+
2026</strong>, in Lugano, Switzerland, <strong>from the 9th to the 13th of February 2026</strong>.
61+
</p>
62+
<!-- <p class="lead">
63+
The USI-SUPSI PhD School in Applied Data Science and Artificial Intelligence organizes a week of advanced training dedicated to the latest statistical, computational, and machine learning methodologies for the analysis of dynamic, relational, and spatiotemporal data,
64+
</p> -->
65+
<p class="lead">The tutorial will take place on Thursday, 12th of February</a>, 9:00-12:00.
66+
</p>
67+
<div class="d-flex flex-column flex-md-row justify-content-between align-items-center"
68+
style="gap: 1.5em">
69+
<a class="btn btn-outline-light" href="https://www.supsi.ch/en/winter-school-in-network-data-science-and-artificial-intelligence" role="button"
70+
target="_blank">Winter school website</a>
71+
<!-- <div>
72+
<img src="./img/log-logo.png" height="182px" class="me-2" />
73+
</div> -->
74+
<a class="btn btn-outline-light" href="https://www.supsi.ch/documents/d/phd-datascience-ai/winter-school-2026-program"
75+
role="button" target="_blank">Winter school program</a>
76+
</div>
77+
</div>
78+
</div>
79+
</div>
80+
</div>
81+
</section>
82+
<!-- Program section-->
83+
<section class="bg-light">
84+
<div class="container px-4">
85+
<div class="row gx-4 justify-content-center">
86+
<div class="col-lg-8">
87+
<h1 class="text-center">Program</h1>
88+
<h5 class="mt-3 mb-2 fw-light"><span class="fw-bold me-2">Part 1</span> Graph deep learning for time
89+
series processing.</h5>
90+
<ol>
91+
<li><strong>Correlated time series</strong><br>
92+
Definition of the problem settings. Time series forecasting.
93+
</li>
94+
<li><strong>Graph deep learning for time series forecasting</strong><br>
95+
Graph-based framework for representing correlated time series. Graph-based models for time
96+
series forecasting.
97+
Similarities to and differences from related settings in time series analysis and temporal graph
98+
learning.
99+
</li>
100+
<li><strong>Spatiotemporal graph neural networks (STGNNs)</strong><br>
101+
Core components of a STGNN model. Recipes and strategies for building STGNNs. The
102+
time-then-space and time-and-space paradigms.
103+
Overview of architectures from the literature.
104+
</li>
105+
<li><strong>Global and local models</strong><br>
106+
Parameter sharing in time series models. Review of the global and local modeling paradigms with
107+
their strengths.
108+
Practical implications in graph-based time series processing. Hybrid global-local STGNN
109+
architectures. Transfer learning.
110+
</li>
111+
</ol>
112+
<h5 class="mt-3 mb-2 fw-light"><span class="fw-bold me-2">Part 2</span> Challenges and tools.</h5>
113+
<ol>
114+
<li><strong>Latent graph learning</strong><br>
115+
Methods to apply the framework when no pre-defined graph is available. Learning graph structures
116+
from collections of time series.
117+
</li>
118+
<li><strong>Scalability</strong><br>
119+
Methods to enable scalability to large sensor networks.
120+
</li>
121+
<li><strong>Dealing with missing data</strong><br>
122+
The problem of missing data. Overview of methods for graph-based multivariate time series
123+
imputation and kriging.
124+
</li>
125+
<li><strong>Model quality assessment</strong><br>
126+
Statistical tools to test the optimality of graph-based predictors. Identification of time-space
127+
regions where predictions can be improved.
128+
</li>
129+
</ol>
130+
<h5 class="mt-3 mb-2 fw-light"><span class="fw-bold me-2">Demo</span> Coding Spatiotemporal GNNs</h5>
131+
<p>Overview of open-source Pytorch libraries for graph-based time series processing. Torch
132+
Spatiotemporal demo.</p>
133+
</div>
134+
</div>
135+
</div>
136+
</section>
137+
<!-- Organizers section-->
138+
<section>
139+
<div class="container px-4">
140+
<div class="row gx-4 justify-content-center">
141+
<div class="col-lg-8 text-center">
142+
<h1>Organizers</h1>
143+
<p class="lead">This tutorial is organized by the <a href=”{{site.url}}” class="fw-normal"
144+
target="_blank">GMLG Research
145+
Group</a> within the Swiss AI Lab <a href="https://idsia.ch" class="fw-normal"
146+
target="_blank">IDSIA</a> and <a href="https://usi.ch" class="fw-normal"
147+
target="_blank">Università della Svizzera italiana</a>,
148+
with substantial contributions of
149+
{% assign ac = people | where: "id", "acini" | first %}
150+
<a href="{{ac.links.website}}"><strong>Andrea Cini</strong></a>
151+
and
152+
{% assign im = people | where: "id", "imarisca" | first %}
153+
<a href="{{im.links.website}}"><strong>Ivan Marisca</strong></a>
154+
to the development of the content.</p>
155+
156+
{% assign people = site.data.people %}
157+
<div class="row g-4 my-4 justify-content-center align-items-center text-center text-md-start">
158+
</div>
159+
<div class="row g-4 my-4 justify-content-center align-items-top text-center">
160+
{% assign dz = people | where: "id", "dzambon" | first %}
161+
{% include people_item.html person=dz hide_description=true col_md=3 %}
162+
</div>
163+
</div>
164+
</div>
165+
</div>
166+
</section>

0 commit comments

Comments
 (0)