On Wednesday June 15, 2022, we hosted a meeting with Helmholtz delegation from Germany.
During their visit we presented our research center and learned about theirs, had an interesting discussion, and 3 researchers from our center presented their work.
We hope it will be the start of a fruitful collaboration with the centers of Helmholtz.
14:30 – Prof. Mor Peleg
Title: Introduction to the Data Science Research Center (DSRC). Prof. Mor Peleg, Dept. of Information Systems and Head of the DSRC
Mor Peleg holds a BSc and MSc degrees in biology and PhD (1999) in Information Systems, from the Technion and post-doctoral studies (1999-2003) at Stanford’s Medical Informatics program. She is Full Professor at the Department of Information Systems, University of Haifa, which she joined in 2003, Head of the University of Haifa’s Data Science Center, former Chair of the Department of Information Systems, and founder and past Chair of the BSc Data Science program. She is Editor-in-Chief of the Journal of iomedical Informatics, and is Fellow of the American College of Medical Informatics and of the International Academy of Health Sciences Informatics. Prof. Peleg is internationally renowned in the area of clinical decision support, with a focus on using ontologies to integrate patient data with clinical knowledge. She led the large-scale European project MobiGuide, providing sensor-monitoring and evidence-based personalized decision-support to patients any time everywhere. Her current research exploits AI and Big Data for a new model of cancer care.
14:45 – Dr. Itzik Klein
Title: Data-Driven Autonomous Navigation and Sensor Fusion
The purpose of navigation is to determine the position, velocity, and orientation of manned and autonomous platforms, humans, and animals. Obtaining accurate navigation commonly requires fusion between several sensors, such as inertial sensors and global navigation satellite systems, in a model-based nonlinear estimation framework. Recently, data-driven approaches applied in various fields show state-of-the-art performance, compared to rule- or model-based methods. This talk will address data-driven based navigation algorithms, derived at the autonomous navigation and sensor fusion lab that enhance common navigation and estimation tasks. The algorithms include: 1) autonomous underwater vehicle navigation 2) indoor positioning using the smartphone’s inertial sensors 3) adaptive nonlinear filtering with model uncertainty for autonomous vehicles 4) quadrotor dead reckoning 5) data-driven denoising of accelerometer signals.
Itzik Klein received the B.Sc. and M.Sc. degrees in Aerospace Engineering from the Technion – Israel Institute of Technology, Haifa, Israel, in 2004 and 2007, respectively, and a Ph.D. degree in Geo-information Engineering from the Technion – Israel Institute of Technology, in 2011. He is currently a Senior Lecturer (Assistant Professor), heading the Autonomous Navigation and Sensor Fusion Lab, at the Hatter Department of Marine Technologies, University of Haifa. Itzik serves as an IEEE Senior Member and has exceptional experience in navigation systems and sensor fusion. He has worked in the industry on navigation related subjects for more than 15 years at leading companies in Israel, prior joining University of Haifa. His research interests include data-driven based navigation, novel inertial navigation architectures, autonomous underwater vehicles, sensor fusion, and estimation theory.
14:55 – Prof. Doron Chelouche
Title: How Empty is Space? The Case for Astrostatistics and Astroinformatics
Astronomy is the oldest discipline of the natural sciences. With the advance in observing capabilities of ground and space telescopes, astrostatistics and astroinformatics are emerging fields of science. These have the potential to shed light on fundamental scientific questions pertaining to the history of the universe and its constituents, the birth of death of stars, and the properties of planets outside our solar system. In this talk some of the outstanding questions in the field will be reviewed with a particular emphasis on the ability of astrostatistics to detect cosmic dust at the edge of the universe with implications for the emergence of life in the universe.
Doron Chelouche is a professor of physics and vice-dean of research at the faculty of natural sciences. He holds a Ph.D. from Tel-Aviv University (2005). He was the first Israeli to have been awarded a Chandra postdoctoral fellowship, and became a member of the school of natural sciences at the Institute for Advanced Study in Princeton (2005-2008). He later joined the Canadian Institute for Theoretical Astrophysics at the University of Toronto before assuming a position at the University of Haifa in 2009. His research interests are in the fields of astronomy and astrophysics, focusing mainly on the study of supermassive black holes and of diffuse material across space. His research has been supported by multiple grants from the EU (Marie Curie IRG), DFG, Israeli Science Foundation, Planning and Budgeting Committee, and NASA.
15:05 – Dr. Anna Brook
Title: Machine Learning for the Geosciences
Advances in sensor technology reveille new opportunities for identifying and predicting changes in Earth’s ecosystems in response to climate change to understanding interactions among the ocean, atmosphere, and land and provided alternatives to the conventional (broadband, low to medium resolution multispectral) satellite sensors. Hyperspectral airborne sensors e.g. AVIRIS, together with active sensors such as light detection and ranging (LIDAR), Radio Detection and Ranging (RADAR), and Terrestrial Laser Scanning (TLS) have shown benefits for uniquely challenging problems in Geoscience. The developments in Unmanned Aerial System (UAS) technology introduce advances in sensor performance and miniaturization. These systems are capable of carrying multispectral, high spatial resolution spectrometers, LiDAR, and thermal sensors. UAS products have the potential to determine initial and extended fire impacts and offer land managers options for high spatial and temporal pattern recognition and anomaly detection.
Consequently, these big data (multi-sensor spectral data) can no longer be analyzed by any of the conventional methods. Geosciences need to process large and rapidly increasing amounts of data to provide more accurate, less uncertain, and physically consistent inferences in the form of prediction, modeling, and understanding the complex Earth system. To get the most out of the rapid growth and diversity of Earth system data, we face two major tasks in the coming years: extracting knowledge from the data, and deriving models that learn much more from data than traditional data assimilation approaches. Advanced machine learning (ML) proposes methods that optimize their performance iteratively by learning from the data. These approaches demand detailed information, thus spectral reduction is not necessary and even undesirable. Such methods can be predictive or distinguish between different classes or patterns. Among the main subsets of ML, applications of genetic and biological programming in the geoscience and remote sensing domain are very new and restricted to a few areas.
Multi-source, multi-scale, high-dimensional, complex Spatio-temporal relations, including non-trivial and long-distance relationships between variables. The increasing demand for probability models, which are more advanced classifiers, has been met with alternatives: Artificial Neural Networks (ANNs) 5, Support Vector Machines 6, Wavelet Analysis, Logistic Regression, and decision trees, as classifiers for high dimensional data. ML techniques and recently established deep learning (DL) models have been proposed for spatial, spectral, and temporal imaging data classification. Deep learning is well-positioned to address the multi-source, multi-scale, high-dimensional, complex data challenges, and network architectures, and developed algorithms produce approaches that address both spatial-spectral and temporal context at different scales.
Convolutional neural networks (CNN), at the front of the current state-of-the-art in DL, first achieved successes in the field of image recognition and classification. The retrieval algorithms are actively underway in the field of supervised DL. ML in general and DL in particular offer promising tools to build new data-driven models for components of the Earth system and thus to build our understanding of Earth. Data-driven ML approaches to geoscientific research will not replace physical modeling but strongly complement and enrich it. Thus hybrid modeling approach, coupling physical process models with the versatility of data-driven ML should be our next step.
Anna Brook is the head of the Spectroscopy and Remote Sensing Laboratory at the department of Geography and Environmental Studies, University of Haifa. She received a Ph.D. in Environmental Sciences with a thesis entitled “Reflectance Spectroscopy as a Tool to Assess the Quality of Concrete in situ” from the Porter School of Environmental Studies at Tel-Aviv University in 2011. Her research focus is on questions which drive technological, environmental and social change. Her aim is to bridge the gap between machine learning and geoscience for sustainability and environmental management at the national and international scales Her primary scientific interest is in developing hybrid approaches by coupling physical processes with the versatility of data-driven machine learning to better understand the ecosystems, biodiversity, dynamic processes and environmental responses to stressors, emphasizing sustainability and decision support system development tools.
15:15 – Helmholtz’s Presentation