20^th Mexican International Conference on Artificial Intelligence

Dr. Piero P. Bonissone, Piero P Bonissone Analytics LLC, CEO, USA

PHM Analytics for Industrial AI: Leveraging Model Ensembles

Abstract: In the past, analytic model creation was an artisanal process, as models were handcrafted by experienced, knowledgeable model-builders. More recently, the use of meta-heuristics, such as evolutionary algorithms, has provided us with limited levels of automation in model building and model maintenance. Data-driven models are becoming a commodity, as we have access to a large number of data-driven models by a combination of crowdsourcing, cloud-based evolutionary algorithms, public-domain libraries, outsourcing, in-house development, and legacy models.  In this context, the critical issue will be model ensemble selection and fusion, rather than model generation.

First, we will illustrate the use of model ensembles within the context of Prognostics and Health Maintenance (PHM) of assets such as aircraft engines, medical imaging devices, and locomotives.  We will cover a few case studies of anomaly detection, diagnosis, prediction, and optimization. Then, we will describe the evolution of analytic models in the era of cloud computing, and propose the use of customized model ensembles on demand, inspired by Lazy Learning. This approach is agnostic with respect to the origin of the models, making it scalable and suitable for a variety of applications. We will present results on the fusion of data-driven models for FlyQuest, a GE-sponsored Kaggle competition, in which we crowdsourced the generation of models predicting the estimated runway and gateway arrival (ERA, EGA) over a month of US flights. We will also describe the fusion of hybrid ensembles, combining physics-based and data-driven models to leverage domain knowledge to improve the performance of analytics. Finally, we will highlight some research trends, challenges and opportunities for Machine Learning techniques in this dynamic context of big data and cloud computing.

Bio: Dr. Bonissone is an independent consultant specialized in the use of analytics for Industrial AI applications.  He provides consulting services in machine learning applications and digital transformation, covering project definition and risk abatement, project evaluation, the transition from development to deployment, and model maintenance. A former Chief Scientist at GE Global Research (GE GR), where he retired in 2014 after 34 years of service, Dr. Bonissone has been a pioneer in the fields of analytics, machine learning, fuzzy logic, AI, and soft computing applications. During the last decade at GE GR, he developed multi-criteria decision-making systems to support PHM applications (prescriptive models), ensemble learning to reduce the variance of predictive models, and model lifecycle automation to create, deploy, and maintain analytic models, providing customized performance while adapting to avoid obsolescence.

He is a Life Fellow of the Institute of Electrical and Electronics Engineers (IEEE), a Fellow in the Association for the Advancement of Artificial Intelligence (AAAI), the International Fuzzy Systems Association (IFSA), and a Coolidge Fellow at GE Global Research.  He received the 2012 Fuzzy Systems Pioneer Award from the IEEE CIS. During 2010-15, he chaired the Scientific Committee of the European Centre for Soft Computing.  In 2008 he received the II Cajastur International Prize for Soft Computing from the European Centre of Soft Computing. In 2005 he received the Meritorious Service Award from the IEEE CIS. He received two Dushman Awards from GE GR. He served as Editor-in-Chief of the International Journal of Approximate Reasoning for 13 years. He is in the editorial board of five technical journals and is Editor at Large of the IEEE Computational Intelligence Magazine.  He co-edited six books and has 180+ publications in refereed journals, book chapters, and conference proceedings, with 11,900+ citations, an H-Index of 58 and an i10-index of 175 (by Google Scholar).

He received 74 patents issued by the US Patent Office.  From 1982 until 2005 he has been an Adjunct Professor at Rensselaer Polytechnic Institute, in Troy NY, where he supervised 5 Ph.D. theses and 34 Master theses. He co-chaired 12 scientific conferences focused on Multi-Criteria Decision-Making, Fuzzy sets, Diagnostics, Prognostics, and Uncertainty Management in AI. He has been a member of the IEEE Fellow Committee in 2007-09; 2012-14, and 2016-17. In 2002, while serving as President of the IEEE Neural Networks Society (now IEEE CIS), he was a member of the IEEE Technical Activity Board. He has been an Executive Committee member of NNC/NNS/CIS society in 1993-2012 and 2016-18 and an IEEE CIS Distinguished Lecturer in 2004-14 and 2017-19. He has been a judge in the IEEE Fellow Committee for ten years, since 2007.  Currently he is in his second term as the Vice-Chair of the IEEE Fellows Committee.

Marta Ruiz Costa-jussà, Technical University of Catalunya, España.

Multilingual Machine Translation with Language-Specific Encoder-Decoders: Translation Quality and Gender Accuracy

Abstract: Multilingual Machine Translation usually relies on a shared encoder-decoder. In this talk, we are showing an alternative to this approach based on language-specific encoders-decoders which allows to incrementally add new languages and modalities to the system. We compare both shared and language-specific architectures in terms of translation quality and gender accuracy showing that the language-specific approach is less biased. Further analysis shows that this bias mitigation is related with a higher amount of gender information in the source embeddings.

Bio: Marta R. Costa-jussà is an ERC Researcher at Universitat Politècnica de Catalunya (UPC, Barcelona), member of the TALP and IDEAI, co-leading the MT-UPC Group. She received her PhD from the UPC in 2008. Her research experience is mainly in Machine Translation. She has worked at LIMSI-CNRS (Paris), Barcelona Media Innovation Center, Universidade de São Paulo, Institute for Infocomm Research (Singapore), Instituto Politécnico Nacional (Mexico) and the University of Edinburgh. She has participated in 18 European (including a Marie Curie Action) and Spanish national projects. She has organised 12 workshops in top venues and published more than 100 papers (including high impact journals, e.g. Nature Machine Intelligence, ACM Computer Surveys and Computational Linguistics). She regularly cooperates with companies as a scientific consultant. Currently, she is leading the ERC Starting Grant LUNAR Project and co-leading the Spanish Project of AdaVoice. Recently, she has received two Google Faculty Research Awards (2018 and 2019).

Quantum Machine Learning

Abstract: Quantum computing aims to use quantum phenomena to build computing devices. This technology has great potential to accelerate some algorithmic tasks beyond the capabilities of classical computers. Quantum machine learning (QML) is an area of research that lies at the intersection of quantum computing and machine learning. Recently, QML has attracted a lot of attention, as machine learning is considered one of the flagship applications of future quantum computers. In this talk, we will discuss the basic ideas of quantum computing and how to implement some machine learning methods as quantum algorithms that can exploit future quantum computers. We will also discuss how to use quantum ideas to improve classical algorithms running on current non-quantum computers.

Bio: Fabio A. Gonzalez is a Full Professor at the Department of Computing  Systems and Industrial Engineering at the National University of  Colombia, where he leads the Machine Learning, Perception and  Discovery Lab (MindLab). He earned a Computing Systems Engineer degree   and a MSc in Mathematics degree from the National University of  Colombia, and a MSc and PhD degrees in Computer Science from the  University of Memphis.  His research work revolves around machine  learning and quantum computing with a particular  focus on the representation, indexing and automatic analysis of  multimodal data and, more recently, on quantum machine learning algorithms and applications.

Analyzing human behaviour from multimodal information, new results & challenges

Abstract: Looking at People (LaP) is the field of computer vision dealing with the analysis of human behavior from visual information. Great improvements have been reported in this field for the so-called "obviously visual” behaviors (e.g., gesture recognition, pose estimation, etc.). However, it is only recently that the community is targeting more complex human behaviours that are not visually evident and therefore require additional information and specialized mechanisms for their analysis. In this talk I will describe efforts on the automated analysis of such subconscious human behaviors by using multimodal information. Specifically, I will focus on the tasks of deception detection from videos, facial emotion recognition and presentation attack detection. Data, resources and available solutions to these problems will be briefly described. Associated challenges and open problems will be presented.

Bio: Senior researcher scientist INAOE, Mexico, member of the board of directors of ChaLearn USA, Chair of the IAPR Technical Committee 12  Also, he is a regular member of the Mexican Academy of Sciences (AMC) and the Mexican Academy of Computing (AMEXCOMP) and member of the Mexican System of Researchers Level II (SNI). Since 2017, he is editor of the Springer Series on Challenges in Machine Learning. He has been involved in the organization of several challenges in machine learning and computer vision collocated with top venues, see http://chalearnlap.cvc.uab.es/. He has served as co-editor of special issues in IJCV, IEEE TPAMI, and IEEE Transactions on Affective Computing. He has served as area chair for NIPS/NeurIPS, ICML and has been a member of the program committee of venues like CVPR, ICPR, ICCV, ECCV, ICML, NIPS, IJCNN. He has served as competition chair of NeurIPS2020, FG2020 and ICPR2020, NeurIPS2019, PAKDD2019-2018, IJCNN2019.  His research interests are on machine learning, challenge organization, and its applications on language and vision.

NETDER: An Architecture for Reasoning About Malicious Behavior

Abstract: Malicious behavior in social media has many faces, which for instance appear in the form of bots, sock puppets, creation and dissemination of fake news, Sybil attacks, and actors hiding behind multiple identities. In this talk, I present the NETDER architecture (which takes its name from its two main modules: Network Diffusion and ontological reasoning based on Existential Rules), to address these issues. This initial proposal is meant to serve as a roadmap for research and development of tools to attack malicious behavior in social media, guiding the implementation of software in this domain, instead of a specific solution. Our working hypothesis is that these problems – and many others – can be effectively tackled by (i) combining multiple data sources that are constantly being updated, (ii) maintaining a knowledge base using logic-based formalisms capable of value invention to support generating hypotheses based on available data, and (iii) maintaining a related knowledge base with information regarding how actors are connected, and how information flows across their network. We show how these three basic tenets give rise to a general model that has the further capability of addressing multiple problems at once.

Bio: Dr. Maria Vanina Martinez obtained her PhD at University of Maryland College Park and pursued her postdoctoral studies at Oxford University in the Information Systems Group in Artificial Intelligence (AI) and Database Theory. Currently, she is an adjunct researcher at CONICET as a member of the Institute for Research in Computer Science (ICC, UBA - CONICET) and an assistant professor at the Department of Computer Science at University of Buenos Aires. In 2018 was selected by IEEE Intelligent Systems as one of the ten prominent researchers in AI to watch. Her current research focus is on Knowledge Representation and Reasoning and the social and ethical impact of the use and development of AI based systems.

Uncertainty Quantification in Machine Learning

Abstract: Due to the steadily increasing relevance of machine learning for practical applications, many of which are coming with safety requirements, the notion of uncertainty has received increasing attention in machine learning research in the recent past. This talk will address questions regarding the representation and adequate handling of uncertainty in (supervised) machine learning. A specific focus will be put on the distinction between two important types of uncertainty, often referred to as aleatoric and epistemic, and how to quantify these uncertainties in terms of suitable numerical measures. Roughly speaking, while aleatoric uncertainty is due to inherent randomness, epistemic uncertainty is caused by a lack of knowledge. Going beyond purely conceptual considerations, the use of ensemble learning methods will be discussed as a concrete approach to uncertainty quantification in machine learning.

Bio: Eyke Hüllermeier is a full professor in the Institute of Informatics at the University of Munich (LMU), Germany, where he heads the Chair of Artificial Intelligence and Machine Learning. He studied mathematics and business computing, received his PhD in computer science from Paderborn University in 1997, and a Habilitation degree in 2002. Prior to joining LMU, he spent two years as a Marie Curie fellow at the IRIT in Toulouse, France, and held professorships at the Universities of Dortmund, Magdeburg, Marburg, and Paderborn.

His research interests are centered around methods and theoretical foundations of artificial intelligence, with a specific focus on machine learning and reasoning under uncertainty. He made significant
contributions to reasoning about uncertain dynamical systems, theoretical foundations of case-based reasoning, as well as ranking preference learning, amongst others. He has published more than 400
articles on these and related topics in top-tier journals and major international conferences, and several of his contributions have been recognized with scientific awards. Most recently, he was recipient of the L.A. Zadeh Prize 2019 awarded by the IFSA Society for active scientific leadership and excellence in research.