Skip to content

Commit 4f5d4e0

Browse files
committed
Merge branch 'main' of github.com:Graph-Machine-Learning-Group/graph-machine-learning-group.github.io
2 parents e3a6171 + 2e6da85 commit 4f5d4e0

6 files changed

Lines changed: 115 additions & 10 deletions

File tree

_data/people.yml

Lines changed: 18 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -30,8 +30,8 @@
3030
name: Andrea
3131
surname: Cini
3232
group: team
33-
role: Ph.D. Student
34-
description: He is interested in relational inductive biases for the processing of spatiotemporal data.
33+
role: Post-doc Researcher
34+
description: His research focuses on machine learning methods for time series analysis and graph processing.
3535
links:
3636
website: https://andreacini.github.io/
3737
github: andreacini
@@ -44,7 +44,7 @@
4444
surname: Marisca
4545
group: team
4646
role: Ph.D. Student
47-
description: His studies concern the analysis and prediction of irregular spatiotemporal data.
47+
description: He studies the interplay between graph deep learning and irregular spatiotemporal data.
4848
links:
4949
website: https://marshka.github.io/
5050
github: marshka
@@ -57,9 +57,11 @@
5757
surname: Butera
5858
group: team
5959
role: Ph.D. Student
60-
description: Working on object detection and computer vision.
60+
description: His research focuses on model adaptability and generalization for spatiotemporal data.
6161
links:
6262
website: https://usi.to/7hz
63+
github: LucaButera
64+
scholar: b-l-TIUAAAAJ
6365
img: LB.jpg
6466
- id: amanenti
6567
name: Alessandro
@@ -87,6 +89,18 @@
8789
linkedin: tommasomarzi
8890
scholar: uYZ6I-MAAAAJ
8991
img: TM.jpg
92+
- id: vmoretti
93+
name: Valentina
94+
surname: Moretti
95+
group: team
96+
role: Ph.D. Student
97+
description: Her research focuses on time series forecasting in non-stationary environments.
98+
links:
99+
github: valentina-moretti
100+
twitter: Vale__moretti
101+
linkedin: valentina-moretti-
102+
# scholar:
103+
img: VM.jpg
90104
- id: simoscopi
91105
name: Stefano
92106
surname: Imoscopi

_data/publications.yaml

Lines changed: 43 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,31 @@
11
---
2+
- title: 'On the Regularization of Learnable Embeddings for Time Series Processing'
3+
links:
4+
paper: https://arxiv.org/abs/2410.14630
5+
venue: preprint
6+
year: 2024
7+
authors:
8+
- id:lbutera
9+
- G. De Felice
10+
- id:acini
11+
- id:calippi
12+
keywords:
13+
- spatiotemporal data
14+
- graph neural networks
15+
- regularization
16+
- transfer learning
17+
- global-local models
18+
abstract: 'In processing multiple time series, accounting for the individual features of each sequence can be challenging. To address this, modern deep learning methods for time series analysis combine a shared (global) model with local layers, specific to each time series, often implemented as learnable embeddings. Ideally, these local embeddings should encode meaningful representations of the unique dynamics of each sequence. However, when these are learned end-to-end as parameters of a forecasting model, they may end up acting as mere sequence identifiers. Shared processing blocks may then become reliant on such identifiers, limiting their transferability to new contexts. In this paper, we address this issue by investigating methods to regularize the learning of local learnable embeddings for time series processing. Specifically, we perform the first extensive empirical study on the subject and show how such regularizations consistently improve performance in widely adopted architectures. Furthermore, we show that methods preventing the co-adaptation of local and global parameters are particularly effective in this context. This hypothesis is validated by comparing several methods preventing the downstream models from relying on sequence identifiers, going as far as completely resetting the embeddings during training. The obtained results provide an important contribution to understanding the interplay between learnable local parameters and shared processing layers: a key challenge in modern time series processing models and a step toward developing effective foundation models for time series.'
19+
bibtex: >
20+
@misc{butera2024regularization,
21+
title={On the Regularization of Learnable Embeddings for Time Series Processing},
22+
author={Butera, Luca and De Felice, Giovanni and Cini, Andrea and Alippi, Cesare},
23+
year={2024},
24+
eprint={2410.14630},
25+
archivePrefix={arXiv},
26+
primaryClass={cs.LG},
27+
url={https://arxiv.org/abs/2410.14630},
28+
}
229
- title: 'Learning Latent Graph Structures and their Uncertainty'
330
links:
431
paper: https://arxiv.org/abs/2405.19933
@@ -80,7 +107,7 @@
80107
- title: Graph-based Forecasting with Missing Data through Spatiotemporal Downsampling
81108
links:
82109
paper: https://arxiv.org/abs/2402.10634
83-
venue: <i>To appear in</i> International Conference on Machine Learning
110+
venue: International Conference on Machine Learning
84111
year: 2024
85112
authors:
86113
- id:imarisca
@@ -169,7 +196,7 @@
169196
- title: Graph-based Time Series Clustering for End-to-End Hierarchical Forecasting
170197
links:
171198
paper: https://arxiv.org/abs/2305.19183
172-
venue: <i>To appear in</i> International Conference on Machine Learning
199+
venue: International Conference on Machine Learning
173200
year: 2024
174201
authors:
175202
- id:acini
@@ -197,11 +224,12 @@
197224
- relational inductive biases
198225
- graph neural networks
199226
abstract: We focus on learning composable policies to control a variety of physical agents with possibly different structures. Among state-of-the-art methods, prominent approaches exploit graph-based representations and weight-sharing modular policies based on the message-passing framework. However, as shown by recent literature, message passing can create bottlenecks in information propagation and hinder global coordination. This drawback can become even more problematic in tasks where high-level planning is crucial. In fact, in similar scenarios, each modular policy - e.g., controlling a joint of a robot - would request to coordinate not only for basic locomotion but also achieve high-level goals, such as navigating a maze. A classical solution to avoid similar pitfalls is to resort to hierarchical decision-making. In this work, we adopt the Feudal Reinforcement Learning paradigm to develop agents where control actions are the outcome of a hierarchical (pyramidal) message-passing process. In the proposed Feudal Graph Reinforcement Learning (FGRL) framework, high-level decisions at the top level of the hierarchy are propagated through a layered graph representing a hierarchy of policies. Lower layers mimic the morphology of the physical system and upper layers can capture more abstract sub-modules. The purpose of this preliminary work is to formalize the framework and provide proof-of-concept experiments on benchmark environments (MuJoCo locomotion tasks). Empirical evaluation shows promising results on both standard benchmarks and zero-shot transfer learning settings.
200-
- title: Relational Inductive Biases for Object-Centric Image Generation
227+
- title: Object-Centric Relational Representations for Image Generation
201228
links:
202229
paper: https://arxiv.org/abs/2303.14681
203-
venue: Preprint
204-
year: 2023
230+
code: https://github.com/LucaButera/graphose_ocrrig
231+
venue: Transactions on Machine Learning Research
232+
year: 2024
205233
authors:
206234
- id:lbutera
207235
- id:acini
@@ -211,7 +239,16 @@
211239
- relational inductive biases
212240
- image generation
213241
- graph neural networks
214-
abstract: Conditioning image generation on specific features of the desired output is a key ingredient of modern generative models. Most existing approaches focus on conditioning the generation based on free-form text, while some niche studies use scene graphs to describe the content of the image to be generated. This paper explores novel methods to condition image generation that are based on object-centric relational representations. In particular, we propose a methodology to condition the generation of a particular object in an image on the attributed graph representing its structure and associated style. We show that such architectural biases entail properties that facilitate the manipulation and conditioning of the generative process and allow for regularizing the training procedure. The proposed framework is implemented by means of a neural network architecture combining convolutional operators that operate on both the underlying graph and the 2D grid that becomes the output image. The resulting model learns to generate multi-channel masks of the object that can be used as a soft inductive bias in the downstream generative task. Empirical results show that the proposed approach compares favorably against relevant baselines on image generation conditioned on human poses.
242+
abstract: Conditioning image generation on specific features of the desired output is a key ingredient of modern generative models. However, existing approaches lack a general and unified way of representing structural and semantic conditioning at diverse granularity levels. This paper explores a novel method to condition image generation, based on object-centric relational representations. In particular, we propose a methodology to condition the generation of objects in an image on the attributed graph representing their structure and the associated semantic information. We show that such architectural biases entail properties that facilitate the manipulation and conditioning of the generative process and allow for regularizing the training procedure. The proposed conditioning framework is implemented by means of a neural network that learns to generate a 2D, multi-channel, layout mask of the objects, which can be used as a soft inductive bias in the downstream generative task. To do so, we leverage both 2D and graph convolutional operators. We also propose a novel benchmark for image generation consisting of a synthetic dataset of images paired with their relational representation. Empirical results show that the proposed approach compares favorably against relevant baselines.
243+
bibtex: >
244+
@article{butera2024objectcentric,
245+
title={Object-Centric Relational Representations for Image Generation},
246+
author={Luca Butera and Andrea Cini and Alberto Ferrante and Cesare Alippi},
247+
journal={Transactions on Machine Learning Research},
248+
issn={2835-8856},
249+
year={2024},
250+
url={https://openreview.net/forum?id=7kWjB9zW90}
251+
}
215252
- title: Graph Kalman Filters
216253
links:
217254
paper: https://arxiv.org/abs/2303.12021

alari.html

Lines changed: 53 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,53 @@
1+
---
2+
layout: particles_header
3+
title: ALaRI
4+
lead: Advanced Learning and Research Institute
5+
description: ALaRI - Advanced Learning and Research Institute
6+
---
7+
<section class="pb-3">
8+
<div class="container px-4">
9+
<div class="row gx-4 justify-content-center">
10+
<div class="col-lg-8">
11+
<p>The <strong>Advanced Learning and Research Institute</strong> (<strong>ALaRI</strong>) was
12+
established in 1999 at the <a target=_blank
13+
href="https://en.wikipedia.org/wiki/Università_della_Svizzera_italiana">Università della
14+
Svizzera italiana</a> (USI) in <a target=_blank href="https://www.lugano.ch/"> Lugano</a>, to
15+
promote research and education in embedded systems and, later, it was incorporated into the <a
16+
target=_blank href="https://www.inf.usi.ch/en">Faculty of Informatics</a> of the same
17+
university. After almost 20 years of research and premier education in embedded systems, ALaRI has
18+
merged with the <a target=_blank href="https://www.idsia.usi-supsi.ch/"> Dalle Molle Institute for
19+
Artificial Intelligence</a> (<a target=_blank href="https://www.idsia.usi-supsi.ch/">Istituto
20+
Dalle Molle di Studi sull'Intelligenza Artificiale, IDSIA</a>) and was transformed into <a
21+
href="https://gmlg.ch/"> Graph Machine Learning Group</a> (GMLG) concentrating mainly on
22+
research on topics such as graph machine learning, non-stationary environments, spatiotemporal data,
23+
reinforcement learning and dynamical systems.</p>
24+
</div>
25+
</div>
26+
</div>
27+
</section>
28+
<section class="pt-0">
29+
<div class="container px-4">
30+
<div class="row gx-4 justify-content-center">
31+
<div class="col-lg-8">
32+
<h2>Brief History</h2>
33+
<p>ALaRI was founded to fill up a research and educational gap in embedded systems (today better known
34+
as cyber-physical systems or Internet of Things) by ex-Rector of Politecnico di Milano, Prof. <a
35+
target=_blank href="https://en.wikipedia.org/wiki/Luigi_Dadda">Luigi Dadda</a>, Prof. <a
36+
target=_blank href="https://www.deib.polimi.it/ita/personale/details/59838"> Mariagiovanna
37+
Sami</a> (Politecnico di Milano) and Prof. <a target=_blank
38+
href="https://si2.epfl.ch/demichel/"> Giovanni de Micheli</a> (Stanford University, currently
39+
EPFL-Lausanne) with enthusiastic support of industry, especially ST Microelectronics and
40+
Hewlett-Packard who became major supporters of the Institute. Also, companies such as Bosch, csem
41+
and imec, among others, provided support in the form of internships for ALaRI students. The main
42+
source of funding to support ALaRI’s research and infrastructure came from industry, Swiss National
43+
Foundation, European Union, and the university.</p>
44+
<p>The directors of the Institute include Prof. <a target=_blank
45+
href="https://www.deib.polimi.it/ita/personale/details/59838"> Mariagiovanna Sami</a>
46+
(1999-2012, part time from the Politecnico di Milano), Prof. <a target=_blank
47+
href="http://usi.to/iyy"> Miroslaw Malek</a> (2012-2017, full time from the Humboldt University
48+
in Berlin) and then from 2018 Prof. <a target=_blank href="http://usi.to/prz"> Cesare Alippi</a>,
49+
who is now in charge of the new incarnation of ALaRI, the Graph Machine Learning Group. </p>
50+
</div>
51+
</div>
52+
</div>
53+
</section>

assets/img/people/AM.jpg

-6.03 KB
Loading

assets/img/people/VM.jpg

50.6 KB
Loading

index.html

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -23,6 +23,7 @@ <h1 class="fw-bolder">Graph Machine Learning Group</h1>
2323
<div class="col-lg-8">
2424
<h2>About GMLG</h2>
2525
<p class="lead">We are a research team part of the <a href="http://www.idsia.ch/">Swiss AI Lab (IDSIA)</a> at <a href="https://usi.ch/">Università della Svizzera italiana</a>, Lugano, Switzerland. The group is led by Prof. Cesare Alippi.</p>
26+
<p class="lead">The group was previously under the <a href="{{ 'alari.html' | relative_url }}">ALaRI</a> institute.</p>
2627
<a class="btn btn-outline-primary" href="{{site.url}}/people.html" role="button">Meet our team</a>
2728
</div>
2829
</div>

0 commit comments

Comments
 (0)