Nuevo seminario de la serie “Nuevas soluciones con Big Data” organizados por el UC3M-Santander Big Data Institute (IBiDat) donde se presentarán problemas reales en distintos campos y la solución aportada utilizando todos los datos disponibles. Los seminarios intentan ser un punto de encuentro de profesionales y académicos para presentar problemas y analizar posibles soluciones basadas en Big Data.
El siguiente seminario será el jueves 02 de marzo a las 16:00 en formato online y finalizará a las 17:30.
En esta ocasión el seminario tendrá como título Unlocking the Power of Geospatial Analytics y será impartido por dos personas: Miguel Álvarez, que dirige el equipo de Ciencia de Datos en CARTO, equipo especializado en el análisis y modelado de datos geoespaciales y Lucía García-Duarte, que es Data Scientist en CARTO, con experiencia en procesamiento de datos, analítica y modelado espacial.
Resumen de la charla:
«Los datos geoespaciales están en todas partes, desde datos sociodemográficos que describen ciudades, datos de tráfico que nos dan información del estado de las carreteras, datos de movilidad humana que nos dicen qué zonas son más populares, datos de redes sociales, transacciones financieras … Saber explotar estos datos es fundamental para la toma de decisiones en negocios y organismos públicos. En esta charla, Lucía y Miguel presentarán CARTO, una compañía especializada en la explotación de datos geoespaciales. Además, nos mostrarán algunos de estos datos, herramientas que utilizan y cómo trabajan dentro del equipo de ciencia de datos.»
Será en español y se grabará.
Los seminarios son de carácter abierto y gratuito, están organizados por Manuel Cebrián (email@example.com) al que puedes dirigirte para información adicional.
Se agradece confirmación de asistencia rellenando el siguiente formulario https://docs.google.com/forms/d/e/1FAIpQLSfIsYuVLKdZDabbF29fDRhmS10_PuVkmJ_s6_jG9rbxunL9tQ/viewform
We are pleased to present the program of the RECSM Winter Methods School 2023 (March 6-10), which comprises 3 excellent methodological courses.
The courses are open to academics (students, professors, individuals employed by colleges and universities, and other institutions of higher education) and non-academics (consultants and other practitioners from the private sector or national and international organizations) and will be held online and on-site.
The courses are:
- Survey data harmonization across time and space. Instructor: Marta Kołczyńska. March 6th-7th. 8-hour course.
- Questionnaire design for cross-cultural survey research. Instructor: Diana Zavala-Rojas. March 8th-10th. 12-hour course.
- Social network analysis. Instructor: Luca De Benedictis. March 9th-10th. 8-hour course.
- Students: 250€
- Researchers: 400€
- Professionals: 500€
- Students: 300€
- Researchers: 500€
- Professionals: 600€
Universitat Pompeu Fabra (UPF) students will have a 15% discount on the student fee. If you are a UPF student, please contact us (firstname.lastname@example.org) before enrolling in the courses and we will provide you with a discount code.
You can pay via credit/debit card or with a bank transfer.
IFORS 2023 – ABSTRACT ARE WELCOMED THROUGH MARCH 15
23rd IFORS Triennial Conference
Santiago, Chile, July 10 – 14, 2023
Abstract submission system is available online at:
There are over 40 clusters already identified which span a diversity of OR topics. The invited clusters and their chairs are found at https://ifors2023.com/clusters. Abstracts must be written in English with a maximum of 2500 characters; submission deadline: March 15, 2023.
All accepted and presented abstracts will be published online as a monograph with ISBN / DOI identifier numbers.
We have six exciting options for the included Cultural Tours during the conference – possibilities include emphasis on arts, wine regions, and skiing in the Andes! See details of the tours here: https://ifors2023.com/venue/travel
- Margaret Brandeau, Stanford University, USA
- Juan Carlos Muñoz, Minister of Transportation of the Chilean Government and Pontificia Universidad Católica de Chile
- Paolo Toth, Università di Bologna, Italy (The EURO Plenary)
- Andrés Weintraub, University of Chile
- Rene de Koster, Erasmus University, The Netherlands
- Brian Denton, University of Michigan, USA
- Andrea Lodi, Cornell Tech, USA, and the Technion, Israel
- Anna Nagurney, University of Massachusetts Amherst, USA
- Tava Olsen, Melbourne Business School, Australia
- Dolores Romero Morales, Copenhagen Business School, Denmark
- Carleton Coffrin, National Laboratory at Los Alamos Fred Glover, Entanglement, Inc., and Gary Kochenberger, Entanglement, Inc., USA; A Tutorial on Quantum Computing and Combinatorial Optimization
- Nesim K. Erkip, Bilkent University, Turkey: Inventory Problems with Heterogeneous Customers
- Andres Gomez Escobar, University of Southern California, USA: Machine Learning
- Rosa G. González Ramírez, University of Los Andes (Chile) and Eduardo Lalla-Ruiz, University of Twente (The Netherlands): Seaside Planning: Overview and Implementation (2 sessions)
- Kate Smith-Miles, School of Mathematics and Statistics, University of Melbourne, Australia: Stress-testing Algorithms via Instance Space Analysis
The conference webpage https://ifors2023.com/ has details about the hotels with special conference rates along with more logistical information.
Registration opens in March!
Prof. Dr. Alice Smith PC Chair email@example.com
Prof. Dr. Jorge Vera OC Co-Chair firstname.lastname@example.org
Mesa redonda – Organizada por la Real Academia en Colaboración con la Federación Española
de Sociedades de Profesores de Matemáticas
Seminario de Análisis Matemático y Matemática Aplicada
Title: Equivalence results for limiting interpolation spaces and its applications
Speaker: Manvi Grover (Charles University, Prague)
Day: 2nd of March, 2023
Place: Seminario Alberto Dou (Aula 209), Facultad de CC Matemáticas, UCM
Organized by: Instituto de Matemática Interdisciplinar (IMI) y el Departamento de Análisis Matemático y Matemática Aplicada
Seminario de Análisis Matemático y Matemática Aplicada
Title: An introduction to the big and little lip functions
Speaker: Bruce Hanson (St. Olaf College)
Day: 23rd February, 2023
Place: Seminario Alberto Dou (Room 209), Facultad de CC. Matemáticas, UCM
Organized by: Instituto de Matemática Interdisciplinar (IMI) y el Departamento de Análisis Matemático y Matemática Aplicada
|JEDE 6: VI Congreso Científico de Jóvenes en Diseño de Experimentos y Ciencia de Datos. Pamplona, del 5 al 7 de Junio 2023|
|El Instituto de Ciencia de los Datos e Inteligencia Artificial (DATAI) de la Universidad de Navarra te invita a participar en el VI Congreso Científico de Jóvenes en Diseño de Experimentos y Ciencia de Datos (JEDE 6) que, junto con las III Jornadas Científicas de DATAI, se celebrará los días 5, 6 y 7 de junio de 2023 en Pamplona.|
Inscripción: hasta el 15 de mayo de 2023
Plazo de solicitud Beca para estudiantes y envío de trabajos para optar a los premios de jóvenes investigadores: hasta el 31 de marzo 2023: hasta el 20 de marzo 2023
Envío de abstracts: hasta el 30 de abril 2023
Celebración del Congreso: 5, 6 y 7 de Junio de 2023
Más información aquí.
Call for Talks at the Game Theory, Market Design, and Mathematical Economics Cluster
IFORS 2023 (10-14 July 2023, Santiago, Chile, https://ifors2023.com/)
Game Theory, Market Design, and Mathematical Economics session cluster aims to engage those involved in OR/MS/Analytics research and application and promote sharing of experience and best practices from game theory, market design, and mathematical economics across the world.
Cluster topics of interest include all related aspects, such as auctions, matching with preferences, portfolio optimization, stochastic games and control, financial and management accounting, etc.
Some sessions are in preparation already, and you are welcome to submit your abstract there via
https://www.euro-online.org/conf/ifors2023/ using the corresponding Code:
Sessions: Codes: Session Chairs:
Mathematical Economics d633bb1f G.-W. Weber
Game Theory 35b22a83 Ulrike Leopold-Wildburger, G.-W. Weber
Pricing in Non-Convex Markets dd2aed5e Robert Day
Dynamical Systems and Mathematical Modeling in Economics ae4cc26a Katsunori Ano
Operational Research in Financial and Management Accounting 1, 2, 3 78ef05a9, d0496bb5, 60c92e36 Matthias Amen
NN 5b84f506 Aleksandrs Smilgins
New OR methods for Emerging Economies a2aad54f Sadiye Eylul Sadanoglu, G.-W. Weber
Contributed talks can be submitted at https://www.euro-online.org/conf/ifors2023/select_section.
With any questions, please contact any of the Game Theory, Market Design, and Mathematical Economics cluster co-chairs:
Martin Bichler email@example.com, Gerhard-Wilhelm Weber firstname.lastname@example.org
El miércoles 1 de marzo de 2023, se celebrará el siguiente Seminario de Estadística e Investigación Operativa:
Título: «Joint chance-constrained Markov decision processes». Organizado por CITMAga, en formato en presencial.
Ponente: Varagapriya Venkatesan (IIT Delhi)
We consider a finite state-action uncertain constrained Markov decision process under discounted and average cost criteria. The running costs are defined by random variables and the transition probabilities are known. The uncertainties present in the objective function and the constraints are modelled using chance constraints. We assume that the random cost vectors follow multivariate elliptically symmetric distributions and dependence among the random constraints is driven by a Gumbel–Hougaard copula. We propose two second order cone programming problems whose optimal values give lower and upper bounds of the optimal value of the uncertain constrained Markov decision process. As an application, we study a stochastic version of a service and admission control problem in a queueing system. The proposed approximation methods are illustrated on randomly generated instances of queueing control problem as well as on well known class of Markov decision problems known as Garnets.
Fecha: Miércoles 1 de marzo de 2023
Lugar: Salón de Grados de la Facultad de Matemáticas, USC
Más información: «Joint chance-constrained Markov decision processes» (click aquí)
17ª Conferencia de Investigación Operativa y Análisis de la OTAN que tendrá lugar en Laurel (Maryland, EEUU) los días 30-31 de octubre de 2023.
1. The Headquarters Supreme Allied Commander Transformation and the Science and Technology Organization cordially invite the submission of abstracts for papers to be presented at the 17th NATO Operations Research and Analysis (OR&A) Conference. The event is open to all NATO nations, Australia, Austria, Finland, Ireland, Japan, New Zealand, Sweden and Switzerland.
2. At this time, the Programme Committee (PC) is planning for a physical event with limited hybrid options, at Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland, USA. The conference will run from 30-31 October 2023 with an optional training day on 1 November 2023.
3. This year’s conference theme is “Changing character of defence and deterrence: the power of analysis”. Collective defence is at the heart of the Alliance and deterrence is a core element of its overall strategy to prevent war, protect Allies, maintain freedom of action and uphold its values. NATO faces the most complex security environment since the end of the Cold War. Innovations, such as autonomous weapons systems or artificial intelligence are changing warfare, as is the increasing amount of available information. Shifts in the global balance of power, such as the rise of China, are challenging the Alliance’s values, and aggressions, such as Russia’s actions against Ukraine, are threating the security of Allies. These major developments, along with the new NATO Strategic Concept, underscore the need for the Alliance to ensure that its deterrence and defence remains credible and effective. The theme reflects the long-standing practice of Operations Research and Analysis in Defence, tackling ongoing challenges faced by the Alliance and looks to the future to bring new methods to old challenges or well-established methods to future challenges.
4. Interested candidates are invited to submit an abstract (between 150 to 250 words) for consideration no later than 15 April 2023. Candidates are asked to use the abstract submission tool located on the STO Event website:
5. Presented material should be releasable to the public to ensure the broadest possible distribution. If required, material may carry a classification marking up to “NATO UNCLASSIFIED, Releasable to Australia, Austria, Finland, Ireland, Japan, New Zealand, Sweden and Switzerland.
6. The PC will make a selection of submitted abstracts based on relation to conference theme, clarity and readability, technical quality, novelty of contribution and military relevance. All authors will be notified by 20 May 2023. Authors of accepted abstracts will be requested to submit a conference proceedings paper (6-8 page paper, containing early results of research and/or initial findings) or a full journal paper (15-18 page paper, in-depth presentation of more substantial and or significant research findings) by 31 July 2023. Both the conference proceedings paper and the full journal paper will be peer-reviewed for respectively the NATO OR&A Conference proceedings and the STO Review journal. Proceedings and full papers must use, respectively the NATO OR&A conference or the STO Review template. Submissions to the STO Review journal need to be at public releasable level and submissions to the NATO OR&A conference can either be public releasable or limited distribution (marked NATO UNCLASSIFIED, Releasable to AUS, AUT, FIN, IRL, JPN, NZL, SWE and CHE). Submissions of written conference proceedings or full papers are not mandatory but are highly encouraged.
7. Subject to unforeseen restrictions, the PC hopes to offer some form of hands-on training or workshop during the conference to promote continuing education and professional development. Confirmation of topics and details regarding the training opportunities will be announced later.
8. The annual OR&A conference serves as an integral component to the continuous development of the NATO OR&A Community of Interest. It builds upon previous conferences and workshops and assembles esteemed members of the profession from numerous organizations, including NATO commands and agencies, national defence analysis and research organizations, centres of excellence, academia, and industry. Our goal is to provide an exciting program with plenty of opportunities for discussion and debate and provide valuable and relevant training opportunities.
9. Additional information and Frequently Asked Questions may be found on the event website:
Any additional questions or concerns may be directed to the PC Chair: Ms. Sue Collins, email@example.com
The Netherlands Society For Statistics and Operations Research (in Dutch called VVSOR) has an Annual Meeting (AM) committee that is proud to announce the 2023 conference on ‘Statistics & Operations Research for Robust Decision Making”.
On Thursday 23 March 2023 we will have four experienced researchers and speakers give talks:
Peter Grünwald (Research institute for mathematics and computer science in the Netherlands), Frank Pijpers (Statistics Netherlands / University of Amsterdam), Gianluca Baio (University College London) and Julie Rozenberg (World Bank).
The talks start at 12:30 PM CET and we are looking forward to you attending the conference, either in Utrecht or online.
The full program and abstracts are available in the attached PDF that is published in our Dutch VVSOR journal STAtOR. Click https://www.vvsor.nl/articles/vvsor-annual-meeting-2023/ for further information and tickets (free online option available).
Nos complace comunicarte que gracias al incondicional apoyo de ASEPUMA y a la colaboración de la Universidad de Oviedo, el próximo 25 y 26 de Mayo de 2023 se celebrarán las V Jornadas de Big Data y Matemáticas. Te invitamos a que nos acompañes y estaremos encantados de que compartas con nosotros la investigación en la que estás trabajando.
Como siempre, contaremos con la presencia de profesionales de gran prestigio que utilizan las matemáticas y el análisis de grandes bases de datos para optimizar la toma de decisiones en su entorno laboral. Las jornadas se desarrollarán conjugando dos modalidades: presencialmente en la Facultad de Economía de la Universidad de Oviedo (Campus de Gijón) y a distancia a través de medios telemáticos.
Además, se dispone de mesas de investigación en las que se aceptan comunicaciones orales, tanto presenciales como on-line que permitirán dar a conocer las investigaciones que cada uno de nosotros está desarrollando. Para ello, ya está abierto el periodo de envío que finaliza el 5 de mayo a las 23:00.
Dispones de toda la información en la página web de las jornadas que ya se encuentra activa:
Big Data y Matemáticas ASEPUMA – Programa 2023
En esta web, iremos actualizando toda la información relevante de las Jornadas.
Tanto la inscripción a las jornadas como el envío de comunicaciones es completamente gratuita y puedes realizarla en la página del congreso o en el siguiente link:
Big Data y Matemáticas ASEPUMA – Inscripción Jornadas y Envío de Comunicaciones
Te agradecemos de antemano que, si tienes interés en asistir, realices la inscripción lo antes posible para poder controlar el aforo de la modalidad presencial.
Si tenéis cualquier duda o comentario, por favor, no dudes en ponerte en contacto con nosotros, haremos todo lo que esté en nuestra mano para aclararlo (firstname.lastname@example.org; email@example.com).
| ISBA Statistics Seminars |
Tatyana Krivobokova (University of Vienna)
Iterative regularisation methods for ill-posed generalised linear models
February 24 | 11:00 | ISBA C.115
| ISBA Applied Statistics Workshops|
Marie-Anne Colocouris (GSK)
Data Analytics Applied to Vaccines Manufacturing
February 24 | 14:30 | ISBA C.115 + Teams
| ISBA Applied Statistics Workshops|
Vincent Smets (Sciensano)
Antoine Grollinger (UCLouvain)
Exploring the digital food environment in Flanders:
Web scraping of food delivery applications data
February 24 | 16:00 | ISBA C.115 + Teams
| IRES Lunch Seminar|
Lamis Kattan (Georgetown University)
Gender-Based Labor Legislation and Employment:
Historical Evidence from the United States
February 28 | 12:45 | Dupriez D.144
| CORE Brown Bag|
Matteo Bulgarelli (CORE)
The Environmental Implications of Renting the Runway
March 01 | 12:50 | CORE C.035
| UCLouvain Economics Seminar|
Giulio Zanella (University of Bologna)
College education, intelligence and disadvantage:
policy lessons from the UK in 1960-2004
March 02 | 12:45 | Doyen 22
| Louvain Finance Seminar|
Paolo Giudici (University of Pavia)
SAFE Artificial Intelligence in Finance
March 10 | 13:00 | MORE 57
«La transparence des pensions complémentaires»
March 17, 2023 | 10:30-12:30 | Agora 12
|Belgian Financial Research Forum 2023|
April 20-21, 2023National Bank of Brussels
June 20-23, 2023
National Bank of Brussels
To participate, please submit your application to
Prof. Leonardo Iania (firstname.lastname@example.org) by the 1st of April
ENBIS-23: First Call for Papers
ENBIS – the European Network for Business and Industrial Statistics is a platform where statistical practitioners and academic statisticians from Europe and beyond meet, exchange ideas and design new projects. ENBIS sponsors an annual conference and numerous additional events, many of them web-based.
Our annual conference hosts presentations from a wide variety of sectors, ranging from manufacturing to service, from private to public. ENBIS-23 will be hosted by the Department of Applied Statistics, Operation Research and Quality at the Technical University of Valencia (UPV) and will take place at the Nexus – Assembly Hall of the Vera Campus (UPV) in Valencia (Spain), from 10th to 14th September
- The conference sessions are scheduled from 11th to 13th September, with the administrative meetings, the pre- and post-conference courses and workshops taking place on 10th, 13th and14th September.
ENBIS-23 session topics include, but are not limited to, the special areas of interest that exist within ENBIS (formalized in the so-called Special Interest Groups). The non-exhaustive list is:
- • Design of Experiments
• Process Modelling and Control
• Data Science in Business and Industry
• Quality Improvement and Six Sigma
• Statistics in Practice
• Reliability and Safety
• Statistics & Machine Learning/Artificial Intelligence
• Data Mining and Warehousing
• Teaching Business and Industrial Statistics
• Measurement Uncertainty
• Statistics in Industry (e.g. Pharmaceutical, Automotive, Chemical, Food)
• Statistical Computing
• Statistical Modelling
• Data Analytics, Optimisation
• Statistics in Business and Finance
• Statistical Engineering
• Statistical Consultancy
• Time Series Modeling and Forecasting
• Industry 4.0 and Digital Twins
• Epidemiology, statistics in healthcare
• Statistics in Environment and Sustainability
• Case Studies of Applications in Industry
• Interpretable Models for Business and Industry
• Risk Management and Risk Analysis
• Statistical Methods in Image Analysis
The conference will also include a series of special thematic sessions.
The list of keynote speakers for ENBIS-23 includes:
- Pierre Pinson. Imperial College London
- Richard D. De Veaux. Williams College
- ENBIS-23 will also have some pre- and post- conference events in the framework of the conference. More information about these events will be available on the ENBIS-23 site.
We cordially invite you to submit an abstract (maximum of 250 words) by 30th of April 2023, at the latest. To submit your abstract go to https://conferences.enbis.org/e/enbis-23, and click on “Call for abstracts”. If you are not already a member of ENBIS, please join first (note there is no membership fee for individual members of ENBIS).
Should you need any conference-related information which cannot be located on the website, please feel free to contact András Zempléni (email@example.com), Chair of the ENBIS-23 Programme Committee, or the Chair of the ENBIS-23 Organizing Committee Alberto Ferrer (firstname.lastname@example.org).
- You can also send your inquiries to the ENBIS Permanent Office (email@example.com).
Del 15 al 19 de julio de 2024, la Universidad de Sevilla será la sede del 9no Congreso Europeo de Matemáticas (9ECM). Ya está abierta la convocatoria para recibir propuestas para la celebración de congresos satélites a este evento, que está disponible en la página web del Congreso:
Compositional Data Analysis course 2023 at University of Girona
We are delighted to announce that the 10th edition of our one-week Compositional Data Analysis course (the week-CoDaCourse) is scheduled to held in Girona (Spain) from the 3rd to the 7th of July 2023.
Please visit http://www.compositionaldata.com/codacourses.php for further details and registration.
This course is promoted by ECAS (http://ecas.fenstats.eu) and SEIO (https://www.seio.es). Members of the CoDa-Association (https://www.coda-association.org/en/) and SEIO enjoy discounted fees.
Note that SEIO is kindly offering two grants to attend the week-CoDaCourse 2023.
We would appreciate if you please forward this announcement to anyone else in your network that might be interested.
With best CoDa-regards,
(Chair of CoDaCourse 2023)
When: 3-7 July 2023
Where: Girona, Spain
Compositional data (CoDa) consist of multivariate data carrying relative information, such as those referring to chemical or nutritional compositions, time-use behavioural patterns, microbiome species distributions, multiparty electoral data, investment portfolios and similar. The particular features of CoDa as parts of a whole require a coherent approach to statistical analysis that takes into account the very own nature of the data and facilitates meaningful interpretations of the results.
This course introduces to the methodological and practical aspects of the statistical analysis of CoDa, as well as an informal forum for discussion. The lectures review the fundamentals and the current state of the art in the field. The practical sessions focus on the application of CoDa techniques and their interpretation using current software packages. An open discussion session is also planned. The attendants are encouraged to bring their own research problems and data sets for discussion.
Both statisticians and applied scientists dealing with CoDa in any field of knowledge are strongly encouraged to attend. As background it is recommended to have taken a one semester undergraduate course in statistics and mathematics. Basic knowledge of multivariate data analysis will be also beneficial.
The course is organised by the Research Group on Compositional Data Analysis of the Department of Computer Science, Applied Mathematics and Statistics of the University of Girona. The course is accredited by the International Association for Mathematical Geosciences (IAMG) since 2011 and by the CoDa-Association since 2017. Discounted fees are available for students and members of the CoDa-Association and the Spanish Society of Statistics and Operations Research (SEIO).
The course will be held at the premises of the University of Girona, obeying all current regulations to prevent the spread of COVID-19. The number of attendants is limited to 30.
For your spare time, note that Girona is a beautiful and well-connected small city nearby the Mediterranean sea, offering excellent local amenities and possibilities for sightseeing, cultural activities and entertainment; accompanied by first-class gastronomy (https://www.girona.cat/turisme/eng/). The province of Girona is home of the renowned Costa Brava, with astonishing mountain and coastal landscapes at short distance from the capital city.
University of Milan-Bicocca
Fellowships 39th cycle
Five fully funded PhD fellowships in STATISTICS are available within the PhD programme ECOSTATDATA at the University of Milan Bicocca for the 39th cycle.
The PhD program extends over four years (Nov 2023 – October 2027)
The first public call for applications will be out from March 21 to April 19, 2023, and the final ranking of candidates will be published in May 2023.
Participation is open also to candidates who do not have an MSc but expect to obtain their degree before October 31, 2023.
There is no specific restriction on the subject of the degree; this will be evaluated by the examining board.
The PhD program ECOSTATDATA comprises three separate curricula: Economics, Statistics, Big Data and Analytics for Business.
Each track has its own class and course schedule. However there is some overlap to foster interdisciplinary teaching and projects.
The first part of the PhD program (about 40%) is devoted to courses on a variety of Statistics and Data Science topics (Stats curriculum); the remaining part is research-oriented leading to the final dissertation.
A student will typically spend a year abroad during the thesis period in a host institution, and in that case the amount of the fellowship will be increased by 50%.
For more information on the PhD program see https://www.dems.unimib.it/en/research/phd-programme
The application procedure for the 39th cycle will be online soon.
Meanwhile you can get useful information on the documents to submit, the examination process, as well as financial aspects, by checking the previous call (38th cycle) at https://en.unimib.it/education/postgraduates/doctoral-research-phd-programmes/applyingdoctorate/calls-application
On March 6, 16.30 (CET), Ivana Ljubic will give a talk at the Online Seminar Series Machine Learning NeEDS Mathematical Optimization, https://congreso.us.es/mlneedsmo/.
Title: Submodular maximization of concave utility functions composed with a set-union operator
Abstract: Many important applications in machine learning deal with submodular maximization. Indeed, finding a subset of data points to train neural networks, determining which data points to label to train a classifier, identifying the most influential nodes in a social network, or generating a representative set of diverse recommendations, are some examples of problems that can be solved using techniques of submodular maximization.
In this talk we focus on a family of discrete optimization problems asking for the maximization of the expected value of a concave, strictly increasing, and differentiable function composed with a set-union operator. The expected value is computed with respect to a set of coefficients taking values from a discrete set of scenarios. The function models the utility function of the decision maker, while the set-union operator models a covering relationship between two ground sets, a set of items and a set of metaitems. The goal is to find a subset of metaitems (satisfying certain budget constraints) that maximizes the total utility corresponding to the items it covers. It has applications to, among others, maximal covering location, and influence maximization problems.
The problem can be modeled as a mixed integer nonlinear program (MINLP) involving binary decision variables associated with the items and metaitems. We propose a double-hypograph decomposition which allows for projecting out the variables associated with the items by separately exploiting the structural properties of the utility function and of the set-union operator. Thanks to it, the utility function is linearized via an exact outer-approximation technique, whereas the set-union operator is linearized in two ways: either (i) via a reformulation based on submodular cuts, or (ii) via a Benders decomposition. We analyze from a theoretical perspective the strength of the inequalities of the resulting reformulations, and embed them into two branch-and-cut algorithms. The computational results reveal that, on our testbed of instances, the method based on combining an outer approximation with Benders cuts significantly outperforms the other alternatives.
The talk is based on the article:
Coniglio, F. Furini, I. Ljubic. Submodular maximization of concave utility functions composed with a set-union operator with applications to maximal covering location problems. Mathematical Programming, 196(1), 9-56 (2022), 2022
Link to the talk: https://eu.bbcollab.com/guest/3953e5e4f39546019de1969c63656eda
Looking forward to e-seeing you there!
The organizers of the Online Seminar Series Machine Learning NeEDS Mathematical Optimization
Emilio Carrizosa, IMUS-Instituto de Matemáticas de la Universidad de Sevilla
Thomas Halskov, Copenhagen Business School
Kseniia Kurishchenko, Copenhagen Business School
Cristina Molero-Río, IMUS-Instituto de Matemáticas de la Universidad de Sevilla
Jasone Ramírez-Ayerbe, IMUS-Instituto de Matemáticas de la Universidad de Sevilla
Dolores Romero Morales, Copenhagen Business School