Calls 2021

Applications relying on the most modern sensing devices and technologies, also combining complex artificial intelligence tasks are now mainstream. The typical approach to enable intelligent applications is cloud-centric, meaning that the intelligence is hosted in the cloud infrastructure, the sensor data collected by some IoT devices. Shifting intelligence from the cloud to the edge of the network can offer different advantages such as reducing the required bandwidth and latency and also improving users’ privacy. However, reconfigure and deploy an end-to-end processing flow that involves the three aforementioned architectural layers (the cloud, the edge and embedded devices) poses major challenges: many different constraints and trade-offs must be addressed (latency, response time , bandwidth, energy consumption, computational power, computational precision, etc.) The subject of this Ph.D. is to investigate and propose novel optimization (such as e.g. pareto-based optimization) and assessment methodologies to efficiently sample such a complex design space in target application sectors such as home, industry, manufacturing, farming, etc. The reference technological environment covers (but are not limited to) embedded device software engineering (micropython, mbed OS, C languages and dialects, etc.), machine learning frameworks deployable on tiny devices (tinyML, tensorFlow lite, etc.), edge-based frameworks (eclipse Kura, edgeX Fundry, etc.) and cloud-based IoT platforms and services with AI support and components (MS Azure, AWS Greengrass, ThingsBoard, etc.).

Supervised learning is a very popular mechanism to teach machines vision-based tasks and skills. Supervision, however, is a bottleneck for building generic machines that can operate across different contexts, environments and applications, while learning and improving their understanding seamlessly. Ideally, machines should develop their own creative strategies for using the sensed data and their experience to continually learn without humans at their side.
The research activities related to this PhD position will focus on building machine vision algorithms to teach machines to seamlessly understand environments: by exploiting as little supervision as possible, by being independent of the sensor modality being used, and by updating their knowledge when ground truth information becomes incrementally available over time.

The Internet of Things (IoT), including smart objects, wearables, and wireless sensor networks, is becoming a key technology to enable applications and services in several domains. Ultra-low-power embedded devices are pervasive; novel embedded machine learning frameworks have been introduced. Thus, distributing intelligence at the edge is possible, opening exciting research scenarios spanning from novel, innovative hardware for always-on or event-based sensing up to deep learning solutions, federated learning, and continual learning fitting resource-constrained platforms.
Motivated by the challenges of these research scenarios, the research aims to (i) define novel hardware/software approaches to optimize AI at the very edge on energy-efficient embedded devices, in particular for audio processing and/or computer vision; (ii) to explore the potential of distributing and fuse the intelligence in heterogeneous nodes of an IoT (iii) to demonstrate the advantages of the investigated approaches in real-world application scenarios, such as those of smart cities.

Several important problems in modern society, such as pollution and global warming, arise from the inability to achieve cooperation between individuals over a large scale. Recent research is providing a growing evidence of the power of social influence (i.e. peer pressure), in promoting cooperative behavior. This PhD has the goal of developing computational models for modeling human behavior and social interactions and of designing data-driven strategies and incentive schemes for promoting collaboration and cooperation. These approaches will be also compared with gamification strategies in the context of real-world experiments. The work of the student will be in collaboration between two research units of Fondazione Bruno Kessler, MobS (directed by Bruno Lepri) and MoDiS (directed by Annapaola Marconi).

The need to translate audio input from one language into text in a target language has dramatically increased in the last few years with the growth of audiovisual content freely available on the Web. Current speech translation (ST) systems are now required to be flexible and robust enough to operate in different application scenarios and diverse working conditions. On one side, the industry calls for key features like real-time processing, domain adaptability, extended language coverage and the capability to adhere to application-specific constraints (e.g. length or lip-synch constraints in the subtitling and dubbing scenarios). On the other side, the society calls for new efforts towards inclusiveness with respect to specific categories and groups (e.g. gender-sensitivity, customization to the needs of impaired users). Both dimensions (industry and society) face the orthogonal challenges posed by the variability of audio conditions (e.g. background noise, strong speakers’ accent, overlapping speakers). The objective of this PhD is to advance the state of the art in speech translation to make ST flexible and robust to these and other factors.

Machine learning and deep neural networks are extensively and successfully used to process audio on powerful computers, while several problems still need to be solved for porting the technology on low consumption devices with limited resources (both in terms of computation power and memory size).
Research is necessary to reduce the redundancy in neural models to make them portable into the internet of things framework. Along this line of research, the Ph.D. thesis will address the problem of end-to-end neural processing for audio classification, keywords spotting, and privacy-preserving audio processing on resource-constrained embedded devices, considering the trade-off between performance and energy efficiency. Advanced explorative research directions will consider how adapting continual learning techniques to low-power end-devices and if approaches such as collaborative machine-learning without centralized training data (i.e. federated learning) can help in privacy-preserving resource-constrained scenarios.

We are looking for candidates willing to develop novel methodologies based on machine learning, deep learning, pattern recognition and artificial intelligence for information extraction, classification, target detection and change detection in radar and radar sounder images.
The PhD activity will be developed in the context of European Space Agency (ESA) space mission JUpiter ICy moons Explorer (JUICE) in the Jovian system. The candidate will be requested to deal with images acquired from active radar systems including Synthetic Aperture Radar (SAR) images and sub-surface radar sounding data from airborne Earth Observation missions and satellite planetary exploration missions. The activity amins in improving the understanding of subsurface structure and their impact on planetary body climate.
Besides the requirements established by the rules of the ICT school, preferential characteristics for candidates for this scholarship are:
• master degree in Electrical Engineering, Communication Engineering, Computer Science, Mathematics or equivalents;
• knowledge in pattern recognition, deep learning, image/signal processing, statistic/remote sensing/radar.

Gamification principles have proven to be very effective in motivating target users in keeping their engagement within everyday challenges, including dedication to education, use of public transportation, adoption of healthy habits, and so forth. School closures due to the COVID-19 pandemic and thus the sudden change in the management of the students' educational pathways has opened up the need for methods and digital systems able to support teachers in defining educational content and objectives for their classrooms and to keep students engaged in their training path. The goal of this PhD Thesis is to investigate approaches, techniques and tools to design and release educational digital systems for personalized and cooperative learning plans. This will be done exploiting AI techniques for adaptive gamification and will support teachers in the process of defining and monitoring dedicated learning plans for their students. At the same time, it will facilitate learning, will encourage motivation and engagement, will improve student’s participation and cooperation, and will stimulate students to expand their knowledge through dedicated learning plans and personalized feedback.

In the context of dialogues with chatbots it has been shown that endowing neural models with a persona profile is important to produce more coherent and meaningful conversations. Still, the representation of such personas is still very limited, usually based on simple facts. The goal of this PhD Thesis is to make a step forward, trying to grasp more profound characteristics of human personality (such as opinions, values, and beliefs) to drive language generation of conversational agents in multiple domains and languages.

End-2-End models for speech recognition have been steadily improved recently, achieving performance comparable to state-of-the-art systems. This paves the way to the adoption of such solutions also for the extraction of semantically higher information, directly from the raw speech. This allows avoiding approaches based on the combination of speech recognition followed by text processing, with consequent propagation of errors from the intermediate stages. The goal of this thesis is to develop innovative end-to-end systems, eventually based on the transformer model and sequence-to-sequence learning, to address tasks like spoken language understanding, named entity recognition, intent classification and so forth. Ideally, at the end of the doctoral thesis, the candidate would have developed a neural audio processing front end that can be applied to a variety of semantic down-stream tasks.

Machine learning can impact people when it is used to automate decisions in areas such as insurance, lending, hiring, and predictive policing. In many of these scenarios, several works have shown that training data can be unfairly biased against certain populations and groups, for example those of a particular race, gender, or sexual orientation. Since training data may be biased, machine learning predictors must account for this to avoid perpetuating or creating discriminatory practices. This Phd student will work on designing and implementing innovative approaches for fair and explainable machine learning and deep learning algorithms. The selected student will have the possibility of collaborating with the activities of the Human-centric Machine Learning program of the ELLIS society (https://ellis.eu/).

Automated Planning is the problem of synthesizing courses of actions guaranteed to achieve the desired objective, given a formal model of the system being controlled. A class of problems particularly interesting for applications is temporal planning (also called planning and scheduling) where the discrete decisions of "what to do" are coupled with the problem of scheduling (deciding "when to do"). Unfortunately, planning and scheduling techniques suffer from scalability issues and are often unable to cope with the complexity of real-word scenarios, despite the plethora of approaches available in the literature. Recently, efforts such as Deepmind AlphaGO and OpenAI Five hit the headlines, with groundbreaking advancements in the field of reinforcement learning. These techniques are able to automatically learn policies to decide what to do in order to achieve the desired goal. However, they offer no formal guarantee and are not model-based. The research objective of this PhD scholarship is to investigate techniques that combine the formal guarantees offered by automated planning and scheduling with the performance and self-improving capabilities offered by recent advances in deep reinforcement learning to construct self-adaptive planners that can improve over time their performance on specific application scenarios.

In the context of the green deal transition we are looking for candidates willing to develop novel methodologies based on machine learning, deep learning pattern recognition and artificial intelligence for information extraction, classification, target detection and change detection in long and dense timeseries of remote sensing images.
The candidate will be requested to deal with both multi-/hyper-spectral images acquired by passive satellite sensors and Synthetic Aperture Radar (SAR) images acquired from active systems for Earth Observation. Copernicus data acquired by the new ESA Sentinels will be considered. The goal is to design novel methods able to use temporal correlation to model landcover behaviors, changes and trends for a better understanding of phenomena over the past and the future for a better modeling and understanding of climate change.
Besides the requirements established by the rules of the ICT school, preferential characteristics for candidates for this scholarship are:
• master degree in Electrical Engineering, Communication Engineering, Computer Science, Mathematics or equivalents;
• knowledge in pattern recognition, deep learning, image/signal processing, statistic/remote sensing, passive/active sensors.

Multi-Access Edge Computing leverages the network's edge to store and process data and applications locally, and provide fast reactions and efficient use of network and computing resources. Future communication systems and networks, both terrestrial and non-terrestrial, will increasingly require solutions based on AI/ML, virtualization and softwarisation techniques besides traditional communication technologies. The goal of this PhD Thesis is to design and explore novel approaches that leverage on the close cooperation with these domains from an overall system perspective.

The increasing complexity of software systems calls for the development of new methods and tools to design and test software systems characterized by high variability from the point of view of the space of the possible functional configurations, the space of the release architectures, and of the aspects related to dynamic reconfiguration. The goal of this PhD thesis is that of exploring new approaches to the testing, verification and validation of these systems that involve the joint use of model-based and AI based techniques such as planning, machine learning and optimization.

Cyber-physical systems are ubiquitous systems that integrate computation, networking and physical processes. The correctness and dependability of their control software is critical in many high-assurance domains such as space, mobility, and energy. However, their design requires hybrid models that combine continuous dynamics with discrete states and is typically verified with simulation-based testing. The objective of this PhD research is that of advancing the state of the art in the formal verification of hybrid system models, integrating various techniques such as SMT-based model checking, simulation, test-case generation, and fault injection. The candidate will work on both theoretical aspects of the problem, as well as its practical applications in relevant case studies, drawn from the domains of aerospace, automotive, railways, and energy.

3D scene understanding using learning-based approaches is becoming largely employed in several application sectors including industry, automotive, surveillance and cultural heritage. Computational algorithms that are traditionally used to process 2D images, such as object detection, tracking and segmentation, are nowadays being successfully extended to process 3D data (e.g. point clouds). However, traditional 3D approaches rarely use the image content directly for their task, but often rely on mid-level representations (e.g. voxels, sparse point clouds) that disregard the rich context provided by images. Moreover, the availability of 3D data is limited because the effort for annotationing it is greater than its 2D counterpart. The goal of this PhD position is to advance the state of the art about 3D scene understanding by focusing on the aspects of learning-based approaches that can potentially leverage different sensor modalities and the lack of human-annotated data.

Cyber-Physical Systems (CPS) are ubiquitous systems that integrate computation, networking and physical processes. The correctness and dependability of their control software is critical in many high-assurance domains such as space, mobility, and energy. However, their design requires complex component-based hybrid models describing the continuous dynamics of the physical components and the discrete interaction with the control and monitoring components. The objective of this PhD research is that of advancing the state of the art in the formal verification of CPS control design models, integrating various techniques such as SMT-based model checking, contract-based design, abstract interpretation, simulation, test-case generation, and fault injection. The candidate will work on theoretical aspects of the problem as well as its practical applications in relevant case studies, drawn from the domains of aerospace, automotive, railways, and energy.

Conversational agents are designed to interact with users in multiple domains on several topics using natural language. Recently end-to-end systems have started to be tested to fight fake news and hate speech in single turn settings. Still, scaling to full dialogue interactions is a challenging topic, requiring world and domain knowledge together with a deep understanding of argumentative tactics. The goal of this PhD Thesis is to overcome the shortcomings of traditional end-to-end applications in which all components are trained from the dialogs themselves, by incorporating several dialogue, argumentation and domain features.

A clinical case is a statement of a clinical practice, presenting the reason for a clinical visit, the description of physical exams, and the assessment of the patient’s situation. Clinical cases (e.g. discharge summaries, clinical cases published in journals, and clinical cases from medical training resources) provide a very valuable source of information for data-driven technologies aiming at predicting clinical outcomes and patient behaviors. This three-year PhD offers a unique context of a collaboration between FBK, specifically the NLP group, and a Swiss multinational healthcare company, worldwide leader in biomedical research. The research will focus on a number of application oriented tasks, including automatic recognition of clinical entities (e.g. pathologies, symptoms, procedures, and body parts, according to standard clinical taxonomies such as ICD-9, ICD-10 and SNOMED-CT); detection of temporal information (i.e. events, time expressions and temporal relations, according to the THYME TimeML standard), and factuality information (e.g. event factuality values, assessment of the effect of negation, uncertainty and hedge expressions). Italian will be the major language of clinical cases, although technologies will be experimented on other languages. The goal of the PhD is both to advance the state of the art for clinical case analysis for the Italian language, and to deliver prototype applications, which can be further made operative in real settings (e.g. hospitals). The candidate will have the unique opportunity to explore different domains (Natural Language Processing, Machine Learning, Health & Well-Being) being directly coached by very experienced teammates. The involved PhD will work in an international environment, collaborating with a healthcare company, with worldwide presence. The candidate will work both at FBK (Trento) and at the abovementioned company’s premises (both in Italy and abroad).

Required/Preferred Candidate Skills and Competencies: the candidate should possess basic knowledge on Natural Language Processing and Machine Learning techniques (particularly deep learning architectures). Experience on biomedical data will be a plus. Basic programming skills (e.g. Python) would complete the profile. Proficiency in English is required, basic knowledge of Italian preferable.

Among the low temperature electrolysis processes, two main approaches are extensively documented: alkaline water electrolysis (AWE) and proton exchange membrane electrolysis cell (PEMWE). AWE is a well-established and durable technology yet with many shortcomings being a large footprint, difficulties in handling the liquid alkaline electrolyte, and insufficient response time. Anion exchange membrane water electrolysis (AEMWE) can potentially combine the beneficial features of the PEMWE and AWE technologies, low cost, raw materials that do not raise concerns in terms of supply bottlenecks, electrodes that do not include platinum group metals (PGM), stainless steel porous transport layers (PTL) and bipolar plates (BPP), a compact design, the adoption of feeds based on noncorrosive liquids (low concentration alkali or pure water), and differential pressure operation. However, as of today AEMWE is limited by AEMs exhibiting an insufficient ionic conductivity as well as a poor chemical and thermal stability. The thesis will focus onto the development and testing of innovative cell layout and materials and in parallel on the improvement/optimization of existing AEM WEL cells. A second aspect of the work is the development of a novel concept stack AEMWE, based different geometry of flow field and electrolyte distribution to extend the dynamic range of operation as well as the reduction of gas cross over to achieve an high purity hydrogen output. The key concept is to reduce the voltage and increase the current density. The approach of the work will articulate around the following aspects: design of high surface area cell, developing suitable support material for this approach which needs to be active towards water dissociation, selection of the more performant components and finally validation of the AEM-WEL stack layut best candidate.

Required/Preferred Candidate Skills and Competencies: consistent, diligent, independent, innovative, good command of English. Knowledge of Italian is strongly preferred.

AEM (Anionic Exchange Membrane) technology represents one of the most advanced and promising technology for the low temperature electrolysis process for the production of green hydrogen. The most important advantage is given by the low cost, if compared with other similar technologies, like the PEM based electrolysis. The cost reduction is given mainly by the adoption of non-precious-metal compounds in the catalytic layers. Currently the AEM technology is not fully mature for large commercial applications, due to the gap of the AEM durability and efficiency, compared to the other, mature technologies. The main target of this research activity will be focused on the development of advanced deposition technologies to improve the AEM performances in terms of efficiency and durability. More in particular, the research activity will be focused on electrospinning coatings of catalytic nanofibers and on spray coatings technologies. Dry synthesis technologies based on the use of plasma and chemical-plasma hybrid processes will be applied too, as well as sophisticated surface analysis techniques. Physical vapor deposition, surface plasma functionalization and atomic layer deposition techniques will be employed for the membrane surface and bulk properties modification following as much as possible on-pot and scalable methods. X-ray photoelectron and Auger electron spectroscopies will be used to investigate the chemical properties of the membrane’s surfaces related to the searched functionalities, along with the physical properties and mechanical stability of the membranes which will be studied by means of electrical, mechanical and stress measurements. AEM will be characterized about V-I behaviour and impedance measurements to investigate limiting transfer mechanism and main barriers. The final target and expected outcome will also include a preliminary analysis of the scale up of the coating technology to a mass production level, for commercial electrolyzer applications.

Required/Preferred Candidate Skills and Competencies:
- Catalytic materials;
- Advanced nanomaterials for energy applications;
- Deposition techniques for nanostructured layers and thin films;
- Electrolysis technologies.

Hydrogen is the most promising among the potential green gases, an essential energy carrier to enable a deep decarbonization, for the sectors difficult to abate, such as heavy industry and heavy mobility. Hydrogen indeed must be extracted by water through electrolysis or other materials of biological origin and wastes. One of the most urgent needs to solve is the scaling up of the sector involving several solutions in the way hydrogen is stored, moved, transported in between production and utilization. Beyond compression, one promising direction for some sectors is that of energy carriers, such as liquid hydrogen and liquid organic hydrogen carriers. Among these, ammonia. For its specific characteristics, ammonia could be an ideal carrier in terms of physical and chemical properties, energy density, enabling an efficient logistic and an ideal use in the hydrogen chain. One of the gaps is its synthesis and its direct utilization. This is potentially feasible through innovative technologies, such as reversible Solid Oxide Cells. The PhD will focus on these two dimensions to enable a safe and efficient generation of ammonia in (co)-electrolysis processes through Solid-State Ammonia Synthesis (SSAS) and its utilization in Direct Ammonia Fuel Cells The PhD will focus on both modelling and engineering as well as on experimental and validation activities for cell and short stack based Solid State technology. The activities will include: • Preliminary study on enabling key technologies for the ammonia synthesis; • Engineering study for direct ammonia synthesis, to design and develop both the single components level and the overall integrated system in terms of sizing, Balance of Plant, integration layout and controls; • Demonstration on lab scale of ammonia synthesis through SSAS process and its utilization in Direct Ammonia Fuel Cells based on Solid oxide and/or Proton Conductive Ceramic technologies.

Required/Preferred Candidate Skills and Competencies:
- Competences in energy engineering;
- Knowledge of the Hydrogen chain;
- Knowledge on conversion processes using both Electrolyser and fuel cells;
- Lab training in use of hydrogen related compounds, including hydrogen carriers.

This PhD project is focused on using deep learning approaches, computer vision and photogrammetry for understanding visual scenes, in particular related to urban environments and people’s behaviours. The project aims at designing novel deep neural network architectures able to exploit multiple sources of data efficiently and to detect people's behaviours, objects, vehicles, in crowded environments such as streets, squares, malls, etc. For example, the project might involve street-view imagery in conjunction with satellite imagery, 3D data, etc. to predict urban outcomes and people’s behaviour. In addition, the candidate will work on generative models (e.g. Generative Adversarial Networks) to augment training data on urban and behavioral patterns (e.g. people movements), which have to be realistic and capture the high diversity of urban forms and lifestyles observed across the globe. Special emphasis will be given to exploit synthetically generated data within GAN architectures. The ideal candidate is strongly motivated to develop machine learning skills focusing on deep learning computer vision and multi-modal approaches. The project will be supervised by Bruno Lepri (FBK), Nicu Sebe (DISI), and Emanuele Strano (MindEarth).

Required/Preferred Candidate Skills and Competencies: Background in computer vision with knowledge on object recognition, image segmentation, deep learning techniques. Interest for applications to visual scene understanding (in particular, urban environments).

Some recent works have shown that the combination of high-resolution satellite imagery and machine learning/deep learning techniques have proven useful for a range of socio-economic and sustainability-related tasks, from poverty prediction to population mapping, from forest and water quality monitoring to the mapping of informal economic activities and settlements, etc. This PhD project aims at designing novel deep learning approaches and novel ways for combining satellite imagery collected at different temporal and spatial resolutions, combining different types of data (for example, optical + radar), and/or combining satellite imagery with other relevant data such as information captured by mobile phones. Moreover, special focus will be dedicated to address applications characterized by the limited amount of reference training information (e.g., property valuation, spatial wealth distribution, exposure to respiratory diseases, etc.). The ideal candidate is strongly motivated to develop machine learning and remote sensing skills focusing on deep learning, satellite imagery and multi-modal approaches, as well as interested in applications to socio-economic and sustainability challenges. The project will be supervised by Bruno Lepri (FBK), Emanuele Strano and Mattia Marconcini (MindEarth).

Required/Preferred Candidate Skills and Competencies: Background in computer vision with knowledge on object recognition, image segmentation, deep learning techniques. Interest for applications to visual scene understanding (in particular, urban environments).

In recent years there has been a growing interest in conversational AI, and a number of conversational systems are now operative in various sectors, including call centres. This situation has made available a huge amount of user-machine interactions, which have a high potential to be used to improve the system performance. As an example, the capacity to detect the emotional content of the conversation would allow the system to respond in a more appropriate way to the user requests. This PhD grant addresses some of the scientific challenges which are behind the interpretation of very large-scale conversational data, including managing noisy data, topic clustering, semi-supervised intent classification, and emotion detection. The resultant of these techniques will be applied to develop empathic chatbots able to model their answers, in real time, on the basis of the human detected emotions expressed during the conversations.

Required/Preferred Candidate Skills and Competencies: Required: Master’s Degree in Science, Computing & Technology, Statistics, Engineering or Mathematics Preferred: Documented experiences in the use Machine Learning techniques applied to real data.

The intellectual property of the research results that will derive from the activities carried out by the doctoral student is owned by the Company.

The Phd candidate will deal with the analysis of olfactory information using digital methods. In particular, s/he will analyse how odors are described in multiple languages, and how smell-related terminology has evolved over time. S/he will also contribute to the development of a system to extract olfactory information from texts, and to the linguistic analysis and evaluation of the system output.

Game-based motivational systems are emerging as an effective tool to engage users and induce a positive change in human behavior. Games introduce goals, interaction, feedback, problem solving, competition, narrative, and fun learning environments, motivational affordances that can increase end-user engagement and motivation. Gamification has gained significant attention especially in educational contexts, where supporting and retaining students motivation is a constant challenge. Although research has demonstrated that collaborative learning benefits a variety of learning outcomes, while also supporting people’s social, emotional, and psychological well-being, most studies on gamification in the educational context focus on individual work and competitive learning activities, exploiting affordances that highlight each student achievements and progression. The goal of this PhD thesis is to investigate the potential of combining motivational gamification mechanics and social and interactive elements of collaborative learning, analyzing the impact in terms of
students’ achievements, engagement, participation and cooperation. The ideal candidate has a background in Computer Science or Cognitive Science. Game design, educational and cognitive psychology, motivation theories, knowledge on designing and conducting experimental studies, experience with quantitative and qualitative data analysis techniques are a plus for the application and should be acquired during the Phd training.

In recent years there has been a growing interest for psychoeducational interventions delivered by means of mobile applications and personal assistants to support self-care of patients, including those coping with chronic conditions. Although the validity of psychoeducation has been proved repeatedly by previous research, the design of effective behavioral intervention technologies for virtual coaching in the area of healthy coping remains a challenge. The aim of the PhD project is to investigate key features of smart coaching solutions for healthy coping interventions that are engaging to use by patients and produce effective outcomes from a clinical perspective. The ideal candidate will be strongly motivated in developing design skills in the field of behavioral intervention technologies and conversational agents for applications in healthcare. The PhD position is offered in co-tutoring between the Digital Health Lab of FBK and DIPSCO.

Computational thinking together with digital competences and AI education are emerging as important topics in the field of education and technology enhanced learning. Although several courses and educational material are being developed, there is still a lack of technology-based personalized and inclusive approaches that might be used to improve teaching and learning practices for such concepts.The aim of this PhD project is to investigate new technology-based approaches to develop computational thinking skills, to improve digital competences and to make AI education accessible. These may include (but not limited to) end-user programming for teachers and students for personalizing and co-creating learning activities. The ideal candidate has a background in Computer Science, Psychology or Cognitive Science. Experience with design of interactive digital technologies, conduction of experimental and in-the-wild studies as well as competences with educational theories are a plus for the application and should be acquired during the Phd training. The PhD position is offered in co-tutoring between the i3 research unit of the Digital Society center of FBK and the Department of Psychology and Cognitive Science.

Redox flow batteries (RFBs) are a promising technology for large scale energy storage. In RFBs power and energy are decoupled: the former depends mainly on the size of the stack while the latter on the size tanks containing the redox active species. This feature make RFBs ideal for economical, large- scale energy storage. However, cost reduction are in order for allowing a widespread diffusion of this technology. The required cost reductions involve two main components of the system: the electrolytes and the stack. Both need to be optimized for enabling a large scale diffusion of RFBs.
Flow batteries are a complex system their design and optimization usually leads to a trade-off between cost and performances (energy and power density, efficiency, cycling life). Cell and stack design is a core task required for the development and upscaling of flow battery systems but redox flow cells models can be very complex due to the multitude of physical phenomena that need to be considered: electric fields, fluid flow, mass and specie transport in different components, electrochemical reactions, heat transfer. All these phenomena need to be considered for building a digital twin of cell and stack and enable to identifying cell-limiting mechanisms, forecasting cell performance and optimizing the design. Despite some commercial software (e.g., COMSOL) can support this activity these tools present a lack of flexibility and serious constraints concerning their use on HPC platform as well as their performances (this also limit their use to small scale cells). Furthermore multiscale models that couple detailed cell models and system level models are not currently available.
For the above mentioned reasons, in this PhD topic we propose to develop a multiphysics-multiscale tool aimed at supporting redox flow cell and stack design and upscaling. This tool will also enable design optimization supported by different algorithms. The selected candidate will be in charge of developing the models extending opensource modelling platforms such as OpenFOAM and integrating optimization tools such as Dakota. The platform will be composed of three different main components tightly connected with each other: 1) Multiphysics cell and stack model (using for example OpenFOAM), 2) System level model based on transient 1D-0D models (using OpenFOAM, OpenModelica or python) 3) optimization tool. The outcome from the multiphysics cell model will be used either as input or for computing the parameters required by the system level model. Different type of optimization will be developed based on the final objective.
The simulation models will be validated with experimental data from known chemistries and representative prototypes, and show how new chemistries can be explored. The candidate will be in charge of developing and implementing the physical models, validating the models based on experimental data, integrating different model for building a multiscale tool and integrating the optimization algorithms in the work flow to enable design optimization. In order to enable a strong cross-contamination of ideas and experience, we propose that the candidate will also support the experimental activities related to the validation of redox flow cells with known and new chemistries.

In the last years, mobile mapping systems, such as hand-held devices, drones and ground vehicles equipped with active or passive sensors, have been widely used for precise 3D data generation based on advanced geo-referencing solutions.

According to specific scenario and application requirements and constraints, different techniques can be adopted to tackle positioning and navigation tasks, e.g. solutions based on global navigation satellite system (GNSS), ultra-wide band (UWB) transceivers and 5G mobile network technologies.

The research should investigate different geo-referencing solutions when using mapping platforms, trying to understand potentials and limitations, feasibility and needs in typical geomatics scenarios, such as mapping of urban and forestry areas, precision farming, hazard monitoring, indoor mapping, heritage documentation.

Some topics that should be considered in the research are:

- the forthcoming Galileo high accuracy service, which should be investigated and exploited at the best of it state of implementation;

- the fusion of data from multi-frequency and multi-constellation GNSS receivers and AI-enabled stereo cameras or stereo optical and depth sensors embedded in smart phones;

- alternative positioning solutions such as those based on UWB and 5G technologies.

The mentioned topics should be investigated in particular for the evaluation of GNSS- based positioning performances and impact on the 3D data in mobile mapping applications, in particular in the photogrammetric field. Positioning technologies such as RTK, PPP and PPP-RTK should be considered along with suitable strategies for the mitigation of GNSS signal degradation and complete temporary loss.

The research is expected to advance the knowledge on the impact of most recent geo-referencing solutions in geomatic mapping based on the photogrammetric principles. High-accuracy positioning is expected to improve the quality of 3D geometric and spatial data and modelling to be used in professional applications. The output of the research will be made accessible through scientific papers.

Interpretability has become a crucial requirement to support translation of reliable AI models into clinical practice. Throughout the graduate course, the candidate will explore how to make a DL model interpretable and reproducible, and she/he will test such methodologies comparing predictive models trained on integrated imaging (CT/PET/MRI or Digital Pathology), omics and clinic data, both publicly available and original, with the final goal of defining a framework or a pipeline leading from the available data to the explainable and repeatable models.

The Ph.D. student will evaluate control measures against different infectious diseases, both retrospectively and prospectively, by developing mathematical models calibrated against observed epidemiological data and informed by other data relevant to the infection under study. The modelling approach will be tailored to the addressed problems and may include compartmental models, generative models, individual-based simulations as well as Bayesian approaches and will include scenario analysis to compare alternative intervention strategies.

Research activity conducted during the Ph.D. will focus on the development of epidemiological models informed by real-world data aimed at investigating the main determinants of the disease spread in humans. Envisioned approaches range from the study of mechanistic models mimicking the spatio-temporal transmission dynamics of infectious diseases to the use of bayesian approaches applied to detailed epidemiological records.

Deep neural networks (DNNs) have reached a prominent position amongst machine learning systems [4, 1] due to an increasing experimental evidence of their flexibility, expressivity, and effectiveness in addressing functional approximation problems. At the same time, a complete and satisfactory mathematical theory, explaining for instance how to optimally design and train a DNN on some specific task, is largely missing. It is worth noticing that typical DNNs need millions or billions of trainable parameters for achieving such outstanding performances, and must execute a proportional amount of arithmetic operations during the inference step. These characteristics make DNNs very demanding in terms of storage, memory, and energy consumption. This creates an obstacle towards their deployment on low-memory and low-power architectures such as embedded devices and microcontroller units (MCUs). For this reason, the use of quantized neural networks (QNNs), that is, of specific DNNs whose parameters take values in small, finite sets and whose activation functions have a finite range, allows to represent the operands with fewer bits with respect to standard DNNs, thus providing significant benefits with respect to digital hardware constraints [10, 5]. In this sense, QNNs have an enormous potential of applications, for instance in biomedical and environmental research, where in-situ analyses of real-time data could be effectively performed by low-powered & portable devices. At the same time, the lack of a rigorous mathematical theory in support of QNNs is even more striking, also due to the non-differentiability properties of such networks, so that a better comprehension of the mathematics involved seems crucial for their effective application. The objectives of the PhD project will be of both theoretical and applicative nature. The theoretical part of the project will consist in studying some mathematical aspects of the theory of DNNs, including: • techniques for layer-wise regularization & approximation of QNNs; • expressivity properties of suitable classes of QNNs;

Neuroimaging methods, like structural, functional, and diffusion MRI as well as MEG/EEG can be used to investigate the anatomical and functional connectivity of the brain. In this project, the candidate will pursue research on machine learning methods for neuroimaging data to study and characterize brain connectivity, with applications to longitudinal studies and clinical practice. The ideal candidate should have a mixed background in neuroimaging techniques and numerate disciplines, like computer science, engineering, physics, or mathematics. This project is in collaboration with the Division of Neurosurgery, S. Chiara Hospital, Trento (IT).

Quantum Chromodynamics (QCD) is the theory of the strong nuclear forces. At ultrarelativistic energies the degrees of freedom are quarks and gluons and their interactions can be calculated with weak coupling methods. For sufficiently high energies, the gluon density becomes large leading to strong non-linear effects whose description is the goal of the Color Glass Condensate (CGC) effective theory. It is important to apply the latter for studying observables in the forthcoming Electron Ion Collider (EIC).

We propose to investigate the optimization of quantum simulations on trapped-ion quantum processors. The Ph.D. candidate will explore the use of quantum optimal control techniques to tailor ‘analog’ gates at the laser pulse level, as well as the optimization of ‘digital’ quantum circuits built on predetermined primitive gates. The study will identify the most effective methodology to translate near-term trapped-ion quantum computing into meaningful quantum simulations of microscopic systems.

The Limadou project gathers some Italian institutions participating in the China Seismo Electromagnetic Satellite (CSES) mission. CSES consists of a constellation of satellites, designed to pursue the deepest campaign of observation of the ionosphere. One of the most important scientific goals of the mission is to look for correlations between transient phenomena in the ionosphere and seismic events. Among payloads, a set of particle detectors is devoted to the detection of charged particles trapped in the Van Allen Belts, to monitor the solar activity and to measure galactic cosmic rays of very low energy. The APP group of the Physics Department in Trento looks for candidates to a PhD programme on the analysis of the scientific data from the payloads on board the CSES-01 and those to be launched on board the satellite CSES-02 in 2022. The student will focus on time-series analyses and participate in the development of the event reconstruction software. These studies will be carried out using the most modern machine learning techniques for clustering and anomaly detection, using full information from CSES payloads. The activity will be carried out in collaboration with INFN-TIFPA, Fondazione Bruno Kessler and the Institute of the High Energy Physics of Beijing. Candidates familiar with the experimental techniques for the detection of charged particles in space are welcome, as well as basic knowledge of Machine Learning/Deep Learning is recommended.

Chemoresistive gas sensors are high-performance sensing devices with very attractive properties, such as high sensitivity, small size and easy integration into portable eletronics and IoT platform. However, their low selectivity is limiting their full potential in many applications. In order to overcome this shortcoming, the use of artificial intellingence approaches for data data analysis is the most promising. Therefore, the adoption of specific sensor arrays equipped with dedicated algorythm turned out to be the successful strategy for the development of optimized sensing platform for medical applications and the indoor air quality monitoring. The PhD project here proposed aims at the investigation of innovative solutions for data analysis of chemoresistive gas sensors, especially for outdoor air quality monitoring and precision agriculture.

Autonomy requires the ability to devise a suitable course of actions, execute the resulting plan, monitor its execution, detect the presence of faults that may have occurred, identify their nature, recover from them, and replan. In addition to achieving the desired goals, planning is required to optimal with respect to multiple criteria, including resource consumption, timings, and tolerance to faults. The problem addressed in this thesis is to define a trustworthy architecture for autonomy, whose correct operation can be formally proved. This shall be carried out by representing planning, runtime monitoring, fault detection, identification and recovery (FDIR), and robustness evaluation in a unique logic-based framework. Satisfiability Modulo Theory will be used to represent with real-time constraints, nondeterminism, resource consumption and temporally extended goals. Within this framework, the algorithms for planning, monitoring and FDIR will be inspired by modern model checking procedures based on incremental abstraction-refinement to deal with infinite state spaces.

Cloud adoption is on the rise and promises to offer many advantages by leveraging economies of scale; at the same time, new security and privacy challenges arise. As an example, consider the protection of data; while in transit and at rest, cryptographic techniques to guarantee confidentiality and integrity are well-understood and readily available for several different use case scenarios in the cloud. The situation is much less clear for data in use, i.e. during computation, although it is fundamental for trusting cloud service providers without taking for granted their unsupported claims about security assurances especially when sensitive (e.g., healthcare or financial) information is being processed. To achieve this advanced level of data protection, it is fundamental to design and prove the security of technical enforcement mechanisms of confidentiality and integrity policies in Trusted Execution Environments. For the usability of these mechanisms, fundamental security services (including key management and attestation) must be developed, their security and risk level formally assessed, and their deployment automated. The research work to be conducted during the thesis aims to make significant contributions to developing methodologies, automated techniques and tools to assist the development of fundamental services for confidential computing solutions in the cloud with a focus on key management, attestation, and their integration with identity management solutions for both users and machines to establish a root of trust with high assurance. Applications of interest for the integration of foundational services range from confidential AI, databases, and analytics to confidential ledgers and multiparty collaboration of dataset owners.

Deep neural networks are being extensively and successfully used in a variety of audio processing tasks. However, neural solutions typically are computationally eager, in particular during training, and are not suitable for low consumption devices with limited resources (both in terms of computation power and memory footprint). TinyML (or tinyAI) is a branch of neural machine learning that investigates strategies to make AI portable on constrained platforms as those typically available in internet of things frameworks. Along this line of research, the Ph.D. thesis will address the problem of end-to-end neural processing for audio classification, keywords spotting, and privacy-preserving audio processing on resource-constrained embedded devices, considering the trade-off between performance and energy efficiency. Advanced explorative research directions will consider how adapting continual learning techniques to low-power end-devices and if approaches such as collaborative machine-learning without centralized training data (i.e. federated learning) can help in privacy-preserving resource-constrained scenarios.

Integrating formal verification and testing for parametric software product lines

The increasing complexity and configurability of systems requires the development of new methods and tools to design and test parametric software systems and product lines characterized by variability from the point of view of the space of the possible functional configurations, the space of the deployment architectures, and of the aspects related to their dynamic reconfiguration at run-time. This Ph.D. thesis aims at defining novel approaches to the testing, verification and validation of this class of systems exploiting the integration of formal verification techniques and software testing approaches. On the one hand, formal techniques have the potential to analyze and cover all the behaviors of a software system; on the other hand, testing techniques are able to ensure adequate levels of coverage also in case of large and complex systems. The challenge here is the search for new directions and methods to interleave the two approaches in the specific case of large parametric systems, in order to optimize the balance between the adequacy of the coverage of the system's behaviors and the effort of the overall verification activity.

Reconfigurable and trustworthy pandemic simulation

Simulation tools are fundamental to predict the evolution of pandemic, and to assess the quality of counter-measures, e.g. the effect of travel restrictions on the spread of the coronavirus. However, they come with two fundamental requirements. The first is the need for a fast reconfiguration of the simulation, in order to be able to describe the mutating scenarios of the pandemics. The second is the ability to produce correct and explainable results, so that they can be trusted and independently validated. The topic of this research is to devise a model-based approach that is able to represent at a high-level the features of a generic pandemic, from which an efficient simulator can be produced. By means of formal methods, the results of the simulation are guaranteed to be correct by construction, with proofs that can be properly visualized and independently checked. The activity will be carried out as a collaboration of the Center for Health Emergencies (https://www.fbk.eu/it/health-emergencies/), that played a major role during the ongoing pandemics, and the Center for Digital Industry (https://dicenter.fbk.eu/), a leading centers in model-based design.

Condition monitoring and predictive maintenance of complex industrial systems: Model-based reasoning meets Data Science

The advent of Industry 4.0 has made it possible to collect huge quantities of data on the operation of complex systems and components, such as production plants, power stations, engines and bearings. Based on such information, deep learning techniques can be applied to assess the state of the equipment under observation, to detect if anomalous conditions have arised, and to predict the remaining useful lifetime, so that suitable maintenance actions can be planned. Unfortunately, data driven approaches often require very expensive training sessions, and may have problems in learning very rare conditions such as faults. Interestingly, the systems under inspection often come with substantial background knowledge on the structure of the design, the operation conditions, and the typical malfunctions. The goal of this PhD thesis is to empower machine learning algorithms to exploit such background knowledge, thus achieving higher levels of accuracy with less training data.

Planning and scheduling with time and resource constraints for flexible manufacturing

Many application domains require the ability to automatically generate a suitable course of actions that will achieve the desired objectives. Notable examples include the control of truck fleets for logistic problems, the organization of activities of automated production sites, or the synthesis of the missions carried out by unmanned, autonomous robots. Planning and scheduling (P&S) are fundamental research topics in Artificial Intelligence, and increasing interest is being devoted to the problem of dealing with timing and resources. In fact, plans and schedules need to satisfy complex constraints in terms of timing and resource consumption, and must be optimal or quasi-optimal with respect to given cost functions. The Ph.D. activity will concentrate on the definition of an expressive, formal framework for planning with durative actions and continuous resource consumption, and on devising efficient algorithms for resource-optimal planning. The activity will explore the application of formal methods such as model checking for infinite-state transition systems, and Satisfiability and Optimization Modulo Theories, and will focus on practical problems emerging from the flexible manufacturing domain.

Meta-learning for efficient 3D representations

Learning-based algorithms for 3D object description, recognition and retrieval suffer from lack of annotated data, computationally inefficient processing pipelines and poor generalisation ability across different application domains, such as robotic manipulation, automotive, and augmented and virtual reality. All these factors together often hinder the employment of 3D processing pipelines in real-world applications. The goal of this Ph.D. position is to conduct research on novel efficient algorithms for 3D feature representations using deep learning that can effectively replace traditional hand-crafted modules to ultimately improve performance, ease deployment and foster scalability.

The PhD project aims at the investigation of innovative functional nanomaterials towards the selective detection of gaseous compounds. The combined use of microfabrication process, silicon functionalization and solid-state gas sensor technology enables a new breakthrough approach in the development of high-performing gas sensing devices, useful for different applications such as indoor and outdoor air quality monitoring, precision farming and medical screening.

Advanced nanostructured materials can be exploited both to improve the gas sensor performance, and to develop innovative gas monitoring tool, such as monolithic micro-gas chromatographs. These innovative gas monitoring systems requires an interdisciplinary investigation into the operating principles of the gas chromatographic technique. The synergistic effect of the system fluid dynamics, combined with the chemistry of surface interactions, represents the basis of this technology. In particular, the chemistry of heterogeneous solid-gas and/or liquid-gas interactions play a key role in the separation and detection of the analyzed molecules. The adequate functionalization with polar or non-polar stationary phases of the pre-concentrator and of the chromatographic microcolumn, developed by means of silicon microfabrication techniques, is crucial to obtain an adequate separation of the analytes. Likewise, the development and deposition of specific nanostructured materials, which act as an active sensing layer in the sensing platform, enables the detection of previously separated compounds

The aim of this thesis is investigating how to develop, exploit, and combine techniques and approaches borrowed from different research fields, ranging from logic to artificial intelligence, from model checking to statistics, to advance the existing services for process mining.
To this purpose, several are the challenges to be faced in the work as, for example, (i) the capability to represent and exploit secondary aspects for business processes such as data, time, resources; (ii) the capability to align execution information with models; (iii) the capability to manage and reason on extremely large quantity of data (big data);

In knowledge-intensive Digital Society applications it is essential to be able to integrate heterogeneous, large and uncertain knowledge sources and infer new knowledge emerging from such an integration process. We are interested in developing new methods for tackling these problems, both from the foundational and the practical point of view. We identify two possible directions: 1) Ontology-mediated transformation of knowledge structures for efficient migration
We aim at developing solutions for migration of data between heterogeneous representation schemas: an ontology is used as an abstract model of the domain of data; ontology aligned schemas will provide data access as a Virtual Knowledge Graph.
We are interested in automatizing ontology mappings from/to resources and logic-based generation migration rules across resources. 2) Reasoning with weighted information to handle context and exceptions. We recently considered a solution (justifiable exceptions) for reasoning over ontological knowledge that allows for exceptional instances. However, in real-world data it is often useful to quantify the degree of exceptionality or probability of an axiom. We are thus interested in studying defeasible reasoning in (contextualized) description logics by integrating numerical reasoning as well as novel methods of explainability in AI.

This PhD grant will exploit cross-disciplinary competences in three areas, i.e., deep learning, argumentation mining and conversational AI, to support a broader and innovative view of explainable AI. The goal is to advance the state of the art in explanatory dialogues through the capacity to automatically detect the quality of an argument. The grant is related to the ANTIDOTE project (Argumentation-Driven explainable artificial intelligence for digital medicine), which fosters an integrated vision of explainable AI, where low level characteristics of the deep learning process are combined with higher level schemas proper of the human argumentation capacity. The research will focus on a number of deep learning tasks in the medical domain, where the need for high quality explanations for clinical cases deliberation is more critical than in other domains.

As demonstrated by recent events, telecommunications networks’ importance for economic activity and simple human communication cannot be understated. However, while networks held up remarkably well despite a near-doubling of their expected load, their resiliency is not infinite, and their role as information carriers makes them useful targets for malicious actors, from crooks to state-sponsored agents. The advent of programmable ASICs offers a unique opportunity for researchers to observe and customize network device behaviors at a level of detail and time resolution unthinkable with traditional approaches. These, in turn, enable the development and application of new or previously unsuitable strategies for on-the-fly fault detection, isolation and recovery policies, including aspects such as detailed timing and direction of error propagation and intrusion detection and isolation.
This thesis combines the topics of programmable networks and advanced fault and intrusion detections and recovery/isolation, with the aim of improving the resiliency of large-scale telecommunications networks against both failures and targeted attacks.

Machine Learning (ML) is nowadays a consolidated technology embedded in various domains of computer science and information technology. In the recent past years, ML has revolutionised cybersecurity applications, with excellent results in various application areas such as: encrypted traffic classification, intrusion detection and prevention, anomaly detection in industrial control systems, identification of malicious software (or malware), among others.
One important research subfield of ML is called Explainable Machine Learning, which relates to understanding the ML model behaviour by means of various techniques such as feature importance scores, influential training data, etc,. Given the complexity of some black-box ML models, it is inherently difficult to understand why they behave the way they do.
Understanding how a ML model works and how it takes its decisions is paramount in network security. Indeed, the ability to understand why an event is classified as benign or malicious by an ML-based intrusion detection system allows the ML practitioner to take the necessary counteractions to reduce false positive and false negatives rates, and to make the system more robust to Adversarial Machine Learning attacks.

The objective of this thesis is to perform fundamental research in the field of ML explainability (understanding how ML algorithms reason their outputs) and to propose novel tools and methodologies for ensuring good performance of ML-based security systems under various working conditions.

Space and avionics systems are reaching an unprecedented degree of complexity. The process of safety analysis attempts to characterize the likelihood of faults and failures, and to assess the effectiveness of the adopted mitigation measures. Unfortunately, traditional techniques are becoming ineffective, unable to deal with large-scale systems. This thesis will investigate novel methods for safety analysis, based on the adoption of formal models of system and software (nominal and faulty) behaviors. Particularly interesting are the analysis of timing aspects in the propagation of multiple faults to failures and errors, the ability to explain the causality of propagation, and the definition of techniques for on-the-fly fault detection, isolation and recovery policies.

The Internet of Things (IoT), including smart objects, wearables, and wireless sensor networks, is becoming a key technology to enable applications and services in several domains. Ultra-low-power embedded devices are pervasive; novel embedded machine learning frameworks have been introduced. Thus, distributing intelligence at the edge is possible, opening exciting research scenarios spanning from novel, innovative hardware for always-on or event-based sensing up to deep learning solutions, federated learning, and continual learning fitting resource-constrained platforms.
Motivated by the challenges of these research scenarios, the research aims to (i) define novel hardware/software approaches to optimize AI at the very edge on energy-efficient embedded devices, in particular for audio processing and/or computer vision; (ii) to explore the potential of distributing and fuse the intelligence in heterogeneous nodes of an IoT (iii) to demonstrate the advantages of the investigated approaches in real-world application scenarios, such as those of smart cities.

Machine learning and deep neural networks are extensively and successfully used to process audio on powerful computers, while several problems still need to be solved for porting the technology on low consumption devices with limited resources (both in terms of computation power and memory size).
Research is necessary to reduce the redundancy in neural models to make them portable into the internet of things framework. Along this line of research, the Ph.D. thesis will address the problem of end-to-end neural processing for audio classification, keywords spotting, and privacy-preserving audio processing on resource-constrained embedded devices, considering the trade-off between performance and energy efficiency. Advanced explorative research directions will consider how adapting continual learning techniques to low-power end-devices and if approaches such as collaborative machine-learning without centralized training data (i.e. federated learning) can help in privacy-preserving resource-constrained scenarios.

In recent years there has been a growing interest for psychoeducational interventions delivered by means of mobile applications and personal assistants to support self-care of patients, including those coping with chronic conditions. Although the validity of psychoeducation has been proved repeatedly by previous research, the design of effective behavioral intervention technologies for virtual coaching in the area of healthy coping remains a challenge. The aim of the PhD project is to investigate key features of smart coaching solutions for healthy coping interventions that are engaging to use by patients and produce effective outcomes from a clinical perspective. The ideal candidate will be strongly motivated in developing design skills in the field of behavioral intervention technologies and conversational agents for applications in healthcare. The PhD position is offered in co-tutoring between the Digital Health Lab of FBK and UNIPD.

Dialogue agents usually work in limited domains with clear and well defined policies, with little adaptation capabilities
to the contextual and social situations. In this scenario it has been shown that endowing neural models with a consistent user/machine
profile is important to produce more coherent and natural conversations. Still, the representation of such profiles is very limited, usually based on simple facts. The goal of this PhD Thesis is investigating end-to-end approaches and generation strategies such as guided decoding, in combination with large pre-trained language models and external knowledge, in order to improve the naturalness of
dialogues. During the PhD we will always be mindful of the societal impact of the technologies we develop.

With this phd, we want to investigate how recommendation systems can take advantage of integrating background knowledge expressed in some symbolic form as e.g. logical formulas with standard numeric optimization method, based on machine learning approaches.
In particular we want to investigate how the state-of-the art methods for recommended systems based on embeddings can be complemented with logical reasoning on formulas that expresses structural properties and constraints about properties and relations between users and items.