Thesis & projects proposals
GOOGLE CLASSROOM FOR THESIS' STUDENTS
All my thesis' students are cordially invited to join the google classroom (classroom code mwx76vm), also joinable with this [LINK]. There you will find all the information and the registration form in order to keep in contact. The classroom will also be used in order to send information regarding the weekly meeting and it's meet link, which you can however find in my calendar under the code [OHT].
Tutti i miei studenti sono inviati ad iscriversi alla google classroom (classroom code mwx76vm), anche attraverso questo [LINK]. Lì troverete tutte le informazioni ed il form di registrazione anagrafica che sarà utile per mantenerci in contatto. La classroom sarà anche usata per mandare informazioni riguardo ai meet settimanali e per condividere il meet link, che comunque potete anche trovare nel mio calendario con il codice [OHT]
Here you will find some thesis proposals for students
The topics are available both for undergraduate projects (Laurea) and graduate master project (Laurea Magistrale)
[TA01] Advanced neural networks for anomaly prediction in multidimensional signalsA project with Tales Alenia Space. The student will address the problems of time series prediction and anomaly detection. The domain of interest is the Satellite Telemetry data, that is a multivariate time series, where some variables are sensor readings while other provides some contextual information.
[TA02] Generative adversarial neural for multidimensional signals modeling and simulationA project with Tales Alenia Space. The student will address the problems of time series modeling and simulation. The domain of interest is Satellite Telemetry data, that is a multivariate time series, where some variables are sensor readings while other provides some contextual information.
[PS01] An Artificial Intelligence Approach to attention evaluation in assisted driving systemsRelated to the Hermes(WIRED) project.. The student will address the problems of driver's focus and attention with respect to the street's objects in order to evaluate his ability to drive . The domain of interest is the autonomous driving industry, while both machine learning and image processing algorithms will be applied.
[PS02] A Machine Learning Approach to assess the interlocutor's attention and understanding during robotic interactionsRelated to the Hermes(WIRED) project.. A The student will address the problems of interlocutor's attention and understanding when interacting with a robotic device. The domain of interest is the social robots industry, while both machine learning and image processing algorithms will be applied.
[PS03] A Machine Learning Approach to assess cognitive conditions in childrenA inter-department and inter-university project .The student will address the problems of cognitive testing and mental impairment diangosis on children. The domain of interest is machine learning applied to mental healthcare.
[PS04] A Machine Learning Approach to assess cognitive conditions in elderly peopleA inter-department and inter-university project .The student will address the problems of cognitive testing and mental impairment diangosis on elderly pepople. The domain of interest is machine learning applied to mental healthcare.
[PS05] A cloud-oriented system as an overall support for psychometric evaluationA inter-department and inter-university project . Psychological tests generally requires validation for their standardization, simplification and reorganizations. The student will work on the developement of a unified cloud-based resource for the management and execution of al the task related to psychometric testing, from the creation of a test, to its validation and use.
[PS06] A Machine Learning Approach to adaptive interfaces for mentally impaired childrenA project within the European collaboration HASFRIENDS . The student will address the problems of self-adaptive user interfaces to be used by mentally impaired children. The domain of interest is machine learning applied to mental healthcare.
[PS07] A Machine Learning Approach to adaptive interfaces for mentally impaired eldersA project within the European collaboration HASFRIENDS .The student will address the problems of self-adaptive user interfaces to be used by mentally impaired elders. The domain of interest is machine learning applied to mental healthcare.
[PH01] Artificial Intelligence and Neural Networks based approach for signal denosing in non-linear spectroscopy applicationsA project with INFN . Development of neural networks based algorithms of signal classification and denoising for Raman Spectroscopy and related applications. The thesis has a theoretical-experimental character, in fact it foresees the development through Artificial Intelligence algorithms and Neural Network able to isolate the Raman signal from spurious non-linear contributions.
[PH02] Graph Neural Networks and Deep Learning algorithms for fragmented solid objects classification and recompositionA project with "Ente Parco Archeologico dei Fori Imperiali" . Development of deep learning approaches applied to the reconstruction of solid objects from their fragments. A theoretical-experimental thesis is proposed which involves the development of innovative deep learning algorithms based on Graph Neural Networks for the reconstruction of an object starting from its fragments. In a first phase the theoretical model will be developed, and later the algorithm will be applied to one or more real cases in the context of the reconstruction of ancient artifacts.
[PH03] Development and implementation of deep neural networks on hardware accelerators for ultra-fast inference and real time application in high energy physics.A project within the ATLAS collaboration for LHC at CERN. To exploit the flexibility provided by the FPGA systems, the ATLAS exeperiment at CERN is developing novel algorithms based on both conventional methods and new deep neural network architectures that use quantised weights and activations, and compression techniques. The thesis work will include algorithm development and implementation, synthesis in the FPGA firmware by using state-of-the-art synthesis framework to convert deep neural network models in hardware models for the accelerator, and evaluation of physics performance in terms of efficiency and fake rates and FPGA logic resource occupancy and timing.
[PH04] Explainable AI and simplified model building from deep neural networks as real-time analytical decision support systems.Related to the European project MUCCA on xAI in the CHIST-ERA context. The development and application of of distillation techniques to build simplified explainable models, and use and development of new xAI techniques is useful to understand the decisions made by the deep algorithms. Such a model could be applied in real time (also using FPGAs / ACAPs as use-cases).
[PH05] Real-time deep learning techniques for N-MRI image acquisition and improvementA Medical imaging project with Siemens. Improve the acquisition sequence of N-MRI images by means of a real-time control algorithm based on deep neural networks. The graduate's job will consist in developing and studying the best neural network architectures for the specific purpose, building a demonstrator based on a high-statistic sample of MRI images we have available for which all low-level information has been saved, and then participate in the writing of the article with which the technique is proposed.
[PH06] Deep Neural Networks for N-MRI image processingA Medical imaging project with Siemens. Work on a new technique to improve noise removal in MRI imaging that can work even in cases of low background signal. The graduate's job will consist in developing and studying the best neural network architectures for the specific purpose, building a demonstrator based on a high-statistic sample of MRI images we have available for which all low-level information has been saved, and then participate in the writing of the article with which the technique is proposed.
[SL01] A machine learning based approach for the diagnosis and follow-up of schizophrenia based on MRI-related dataA computational medicine project and A.I with S. Lucia hospital. Work on a new technique to improve classification of MRI-related data for the diagnosis and follow-up of schizophrenia.
[SL02] A machine learning based approach for the diagnosis and follow-up of Alzheimer's disease based on MRI-related dataA computational medicine and A.I. project with S. Lucia hospital. Work on a new technique to improve classification of MRI-related data for the diagnosis and follow-up of Alzheimer's disease.
[SA01] Autonomous Robots as a Resource for Mass Epidemic Sustainability assisting in critical care wards.A medical robotics and AI project with S. Andrea hospital. The goal is to implement robots as acting remote interface for physicians at home in smart-working mode, to perform diagnostic tasks that do not require high precision manual interactions, and therefore decreasing the workload of physically available operators in critical care wards. The secondary goal is to implement an autonomous tool in order to spare a large amount of time that is generally wasted by the caregivers in sanitization operations, tampering with number and duration of visits to patients who are normally left alone for a long time, also lowering the quality of service perceived by the patients, as well as harming them also on the psychological side, with significant fallbacks on the recovery speed.
[SA02] A machine learning based approach for Healthcare Emergency Resources Management during Mass Epidemics.A computational medicine project with S. Lucia hospital. The goal is to implement Machine Learning Algorithms to provide a preemptive planning strategy for intensive care units' accesses in order to enforce middle and long term sustainability in pandemic scenarios. The student will implement machine learning algorithms to predict the bed availability in critical care units, as well as predicting the workload availability due to potential contagions of caregivers and also to improve the emergency management of hospital facilities.
[PR01] A Machine Learning Approach to context-driven privacy enforcementRelated to the GreenTAGS PRIN 2017 project. Internet of Things technologies have already begun to tamper with people’s privacy and discretion. The student will work on machine learning based algorithms jointly with rfid technology to enforce privacy rules using image recognition techniques starting from videos and image-based datasets.
[PR02] A Machine Learning Approach to people's activity recognition and localizationRelated to the GreenTAGS PRIN 2017 project. Internet of Things technologies can help identifying people’s location and activities even in a crowded room. The student will work on machine learning based algorithms jointly with rfid technology to classify people's activity and locations starting from videos and image-based datasets.
[PG01] Unsupervised anomaly detection in industrial image data with autoencodersAutoencoders have been widely used for anomaly detection. A common approach consists in using the reconstruction error, that quantifies the differences between the original and reconstructed samples, as an anomaly score. However simple autoencoders are not well suited for this task, because the feature space could be highly nonlinear and map similar samples to far regions in the latent space. In particular we want the representation to be sensitive to changes in the directions that go from a normal instance to another and at the same time insensitive to changes in the directions that goes from a normal sample to an anomalous one. This can be achieved using more advanced architectures like Contractive Autoencoders or Denoising autoencoders. You are asked to compare these architectures on the mvtec dataset, containing industrial products where anomalies are manufacturing defects.
[PG02] The expressiveness of local aggregation (pooling) in point setsThere are a lot of generalizations of the pooling operator for graphs but little work has been done in its application in point sets. However, point sets can be considered graphs, where the set of neighbors of each point is defined according to the euclidean distance. This interpretation allows us to apply local aggregation techniques devised for graphs on point sets. With this project you are asked to study these pooling strategies and to apply them on point sets. In particular we are interested in the application of local aggregation to solve a specific task, that could be shape classification, counting the number of legs in chairs or tables, counting the number of holes in objects with holes etc. In particular we are interested in how this intermediate representation changes according to the specific task.
[PG03] Prediction of solar wind with Convolutional LSTM NetworksThe solar wind is a flux of charged particles released from the Sun. When those particles are ejected in high numbers and within a certain kinetic energy range, they can damage electronic devices on satellites, as well as on the ground. Both proton density and electron density of the solar wind seems correlated with the solar activity and in particular with the evolution of the active regions. In this project you are asked to devise a Convolutional LSTM network that predicts the proton and electron density of the solar wind in the future looking at the recent history of the active regions of the sun. You are provided with a sequence of images containing a binary mask of the active regions and the time series with the proton and electron densities corresponding to the same periods. We are interested in predicting the proton and electron density at least 2 days in advance, that corresponds to the time it takes for the charged particles to reach the earth.
[PG04] Convolutional features visualization in non euclidean dataConvolutional Neural Networks are capable of solving complex vision problems like object classification, segmentation, etc. by building a hierarchical representation of the data. Shallow layers extract low level features like edges and contours and deeper layers aggregate those features into higher level features. Recently some algorithms have been introduced to extract the features that fire the activation of a particular neuron or a whole layer, but a little work has been done in the application of these techniques to non-euclidean data. In this project you are asked to design similar techniques for graph convolutional networks. You can choose a graph dataset from the Open Graph Benchmark but remember that you must be able to interpret the extracted features.