Merge branch 'main' of github.com:Graph-Machine-Learning-Group/graph-machine-learning-group.github.io

Author: dzambon

_data/news.yml

@@ -1,3 +1,9 @@
+- date: 2025/05
+  text: 'Our papers <a href="https://arxiv.org/abs/2405.19933">Learning Latent Graph Structures and their Uncertainty (Manenti et al.)</a> and <a href="http://arxiv.org/abs/2502.09443">Relational Conformal Prediction for Correlated Time Series (Cini et al.)</a> have been accepted at <strong><a href="https://icml.cc/Conferences/2025">ICML 2025</a></strong>!'
+- date: 2025/03
+  text: 'Our workshop <a href="https://sites.google.com/view/ibrl-workshop/home">Inductive Biases in Reinforcement Learning (IBRL)</a> has been accepted to <a href="https://rl-conference.cc/index.html">RLC 2025</a>! Consider submitting your work.'
+- date: 2025/02
+  text: 'Two new papers accepted at TMLR! <a href="https://arxiv.org/abs/2304.05099">Feudal Graph Reinforcement Learning (Marzi et al.)</a> and <a href="https://arxiv.org/abs/2410.14630">On the Regularization of Learnable Embeddings for Time Series Forecasting (Butera et al.)</a>.'
 - date: 2024/08
   text: 'Our <a href="https://ieeexplore.ieee.org/document/10636792">survey on GNNs for time series</a> has been accepted at <strong>IEEE TPAMI</strong>!'
 - date: 2024/05

_data/publications.yaml

@@ -1,9 +1,9 @@
 ---
-- title: 'On the Regularization of Learnable Embeddings for Time Series Processing'
+- title: 'On the Regularization of Learnable Embeddings for Time Series Forecasting'
   links:
     paper: https://arxiv.org/abs/2410.14630
-  venue: preprint
-  year: 2024
+  venue: Transactions on Machine Learning Research
+  year: 2025
   authors:
   - id:lbutera
   - G. De Felice
@@ -15,16 +15,15 @@
     - regularization
     - transfer learning
     - global-local models
-  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.'
+  abstract: 'In forecasting 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 attempting to prevent the co-adaptation of local and global parameters by means of embeddings perturbation are particularly effective in this context. In this regard, we include in the comparison several perturbation-based regularization methods, going as far as periodically 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.'
   bibtex: >
-    @misc{butera2024regularization,
-      title={On the Regularization of Learnable Embeddings for Time Series Processing}, 
-      author={Butera, Luca and De Felice, Giovanni and Cini, Andrea and Alippi, Cesare},
-      year={2024},
-      eprint={2410.14630},
-      archivePrefix={arXiv},
-      primaryClass={cs.LG},
-      url={https://arxiv.org/abs/2410.14630}, 
+    @article{butera2025regularization,
+      title={On the Regularization of Learnable Embeddings for Time Series Forecasting},
+      author={Luca Butera and Giovanni De Felice and Andrea Cini and Cesare Alippi},
+      journal={Transactions on Machine Learning Research},
+      issn={2835-8856},
+      year={2025},
+      url={https://openreview.net/forum?id=F5ALCh3GWG},
     }
 - title: 'Learning Latent Graph Structures and their Uncertainty'
   links: