[UAI] ICLR 2023 Workshop on Domain Generalization

[Henry Gouk]

[This email originated from outside of OSU. Use caution with links and 
attachments.]

ICLR 2023 Workshop: What do we need for successful domain generalization?


Website: https://domaingen.github.io/


The real challenge for any machine learning system is to be reliable and robust 
in any situation, even if it is different compared to training conditions. 
Existing general purpose approaches to domain generalization (DG) — a problem 
setting that challenges a model to generalize well to data outside the 
distribution sampled at training time — have failed to consistently outperform 
standard empirical risk minimization baselines. In this workshop, we aim to 
work towards answering a single question: what do we need for successful domain 
generalization? We conjecture that additional information of some form is 
required for a general purpose learning methods to be successful in the DG 
setting. The purpose of this workshop is to identify possible sources of such 
information, and demonstrate how these extra sources of data can be leveraged 
to construct models that are robust to distribution shift. Specific topics of 
interest include, but are not limited to:


* Leveraging domain-level meta-data
* Exploiting multiple modalities to achieve robustness to distribution shift
* Frameworks for specifying known invariances/domain knowledge
* Causal modeling and how it can be robust to distribution shift
* Empirical analysis of existing domain generalization methods and their 
underlying assumptions
* Theoretical investigations into the domain generalization problem and 
potential solutions

Submissions are accepted via OpenReview: 
https://openreview.net/group?id=ICLR.cc/2023/Workshop/DG

Submission deadline: February 3, 2023
Author notifications: March 3, 2023
Meeting: May 5, 2023
___
uai mailing list
uai@engr.orst.edu
https://it.engineering.oregonstate.edu/mailman/listinfo/uai

[UAI] CFP: ECML/PKDD Workshop on neuro-symbolic metalearning and AutoML

[Henry Gouk]

*ECML/PKDD Workshop on neuro-symbolic metalearning and AutoML*

This workshop explores different types of meta-knowledge, such as
performance summary statistics or pre-trained model weights. One way of
acquiring meta-knowledge is by observing learning processes and
representing it in such a way that it can be used later to improve future
learning processes. AutoML systems typically explore meta-knowledge
acquired from a single task, e.g., by modelling the relationship between
hyperparameters and model performance. Metalearning systems, on the other
hand, normally explore metaknowledge acquired on a collection of machine
learning tasks. This can be used not only for selection of the best
workflow(s) for the current task, but also for adaptation and fine-tuning
of a prior model to the new task. Many current AutoML and metalearning
systems exploit both types of meta-knowledge. Neuro-symbolic systems
explore the interplay between neural network-based learning and
symbol-based learning to get the best of those two types of learning. While
doing so, it tries to use the existing knowledge as a concrete symbolic
representation or as a transformed version of the symbolic representation
suited for the learning algorithm. The goal of this workshop is to explore
ways in which ideas can be cross-pollinated between the AutoML/Metalearning
and neuro-symbolic learning research communities. This could lead to, e.g.,
systems with interpretable meta-knowledge, and tighter integration between
machine learning workflows and automated reasoning systems.

Main research areas:

   - Controlling the learning processes
   - Definitions of configuration spaces
   - Few-shot learning
   - Elaboration of feature hierarchies
   - Exploiting hierarchy of features in learning
   - Meta-learning
   - Conditional meta-learning
   - Meta-knowledge transfer
   - Transfer learning
   - Transfer of prior models
   - Transfer of meta-knowledge between systems
   - Symbolic vs subsymbolic meta-knowledge
   - Neuro-symbolic learning
   - Explainable and interpretable meta-learning
   - Explainable artificial intelligence

Confirmed invited speakers include:

   - Artur d’Avila Garcez
   
,
   City University of London, UK
   - Bernhard Pfahringer
   
,
 University of
   Waikato, New Zealand

Deadline: 26 June
Website: 
https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fjanvanrijn.github.io%2Fmetalearning%2F2023ECMLPKDDworkshop&data=05%7C01%7Cuai%40engr.orst.edu%7C7676901ecab3497bb60608db6c6a6980%7Cce6d05e13c5e4d6287a84c4a2713c113%7C0%7C0%7C638222976509615172%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=h2kcmASy2FjP3NEi%2FVB7PD2qcX1ke6DW%2FKF4PbHoCS8%3D&reserved=0

Best,
Workshop Chairs
___
uai mailing list
uai@engr.orst.edu
https://it.engineering.oregonstate.edu/mailman/listinfo/uai

[UAI] CFP: Workshop on Meta-Knowledge Transfer/Communication in Different Systems @ ECML-PKDD

[Henry Gouk]

*Workshop Meta-Knowledge Transfer/Communication in Different Systems*

Friday 23rd of Sept. 2022, Grenoble, France
associated with ECML/PKDD 2022

https://janvanrijn.github.io/metalearning/workshop2022

*Objectives*

Meta-knowledge normally captures different aspects of existing (or
potential) solutions, including, for instance: which preprocessing methods
should be used for the given data; which machine learning algorithms are
relevant; which hyperparameters should be considered and how these should
be set; and how all these elements should be combined into useful pipelines
or procedures. The aim of this workshop is to address the issues related to
how (meta-)knowledge can be generated and transferred among different ML
and AutoML systems, so that their joint capability to solve problems can be
enhanced.

*Main research areas:*

   -

   Controlling the learning processes
   -

   Definitions of configuration spaces
   -

   Few-shot learning
   -

   Exploiting hierarchy of features in learning
   -

   Meta-learning
   -

   Conditional meta-learning
   -

   Meta-knowledge transfer
   -

   Transfer learning
   -

   Transfer of meta-knowledge between systems
   -

   Symbolic vs subsymbolic meta-knowledge
   -

   Neuro-symbolic learning and meta-learning
   -

   Explainable and interpretable meta-learning

*Invited speakers*

   -

   Pascal Hitzler, Kansas State University
   Some advances regarding ontologies and neuro-symbolic artificial
   intelligence
   -

   Timothy Hospedales, University of Edinburgh
   Meta-learning for Knowledge Transfer


*Important dates*

   -

   Paper submission deadline: 20 June 2022
   -

   Paper acceptance notification: 13 July 2022
   -

   Early registration deadline: July 22th
   -

   Program and papers will be available online by: 5 Sep 2022


* Accepted formats of submissions*

Original paper track
Authors can submit novel papers that have not been accepted elsewhere. All
submissions should follow the LaTeX Lecture Notes in Computer Science
format, maximal 12 pages.

Poster tack of already published work:
Authors can apply for a poster spot for a paper that has recently (in
2021-22) been published elsewhere. During submission, please send a link to
the already published version of the work, and the peer-review will
determine whether it is a good match based on the topics of the workshop.

*Format of the workshop *

The workshop will last a half a day, as it forms part of a combined
tutorial-workshop. It will include:

   -

   Invited talks
   -

   Short oral presentations
   -

   Poster session
   -

   Panel discussions on
   “Main challenges of communication of (meta)-knowledge among different
   systems”


*Proceedings*

The organizers are planning to prepare formal proceedings. The authors of
accepted papers can decide whether they wish to have their full paper
included or not. In the latter case, publication of a short abstract would
be possible.


*Workshop Chairs*

Pavel Brazdil, Univ. of Porto, Portugal

Jan N. van Rijn, Leiden University, The Netherlands

Felix Mohr, Universidad de La Sabana, Colombia

Henry Gouk, Postdoc. School of Informatics, Univ. of Edinburgh, Scotland
___
uai mailing list
uai@engr.orst.edu
https://it.engineering.oregonstate.edu/mailman/listinfo/uai