Interpretable Convolutional Neural Networks for Decoding and Analyzing Neural Time Series Data

Machine learning is widely adopted to decode multi-variate neural time series, including electroencephalographic (EEG) and single-cell recordings. Recent solutions based on deep learning (DL) outperformed traditional decoders by automatically extracting relevant discriminative features from raw or minimally pre-processed signals. Convolutional Neural Networks (CNNs) have been successfully applied to EEG and are the most common DL-based EEG decoders in the state-of-the-art (SOA). However, the current research is affected by some limitations. SOA CNNs for EEG decoding usually exploit deep and heavy structures with the risk of overfitting small datasets, and architectures are often defined empirically. Furthermore, CNNs are mainly validated by designing within-subject decoders. Crucially, the automatically learned features mainly remain unexplored; conversely, interpreting these features may be of great value to use decoders also as analysis tools, highlighting neural signatures underlying the different decoded brain or behavioral states in a data-driven way. Lastly, SOA DL-based algorithms used to decode single-cell recordings rely on more complex, slower to train and less interpretable networks than CNNs, and the use of CNNs with these signals has not been investigated. This PhD research addresses the previous limitations, with reference to P300 and motor decoding from EEG, and motor decoding from single-neuron activity. CNNs were designed light, compact, and interpretable. Moreover, multiple training strategies were adopted, including transfer learning, which could reduce training times promoting the application of CNNs in practice. Furthermore, CNN-based EEG analyses were proposed to study neural features in the spatial, temporal and frequency domains, and proved to better highlight and enhance relevant neural features related to P300 and motor states than canonical EEG analyses. Remarkably, these analyses could be used, in perspective, to design novel EEG biomarkers for neurological or neurodevelopmental disorders. Lastly, CNNs were developed to decode single-neuron activity, providing a better compromise between performance and model complexity.

Decoding Agc1 deficiency: bioinformatics approaches to unveil the functional implications of Slc25a12 on the transcriptome and epigenome

In the brain, mutations in SLC25A12 gene encoding AGC1 cause an ultra-rare genetic disease reported as a developmental and epileptic encephalopathy associated with global cerebral hypomyelination. Symptoms of the disease include diffused hypomyelination, arrested psychomotor development, severe hypotonia, seizures and are common to other neurological and developmental disorders. Amongst the biological components believed to be most affected by AGC1 deficiency are oligodendrocytes, glial cells responsible for myelination. Recent studies (Poeta et al, 2022) have also shown how altered levels of transcription factors and epigenetic modifications greatly affect proliferation and differentiation in oligodendrocyte precursor cells (OPCs). In this study we explore the transcriptomic landscape of Agc1 in two different system models: OPCs silenced for Agc1 and iPSCs from human patients differentiated to neural progenitors. Analyses range from differential expression analysis, alternative splicing, master regulator analysis. ATAC-seq results on OPCs were integrated with results from RNA-Seq to assess the activity of a TF based on the accessibility data from its putative targets, which allows to integrate RNA-Seq data to infer their role as either activators or repressors. All the findings for this model were also integrated with early data from iPSCs RNA-seq results, looking for possible commonalities between the two different system models, among which we find a downregulation in genes encoding for SREBP, a transcription factor regulating fatty acids biosynthesis, a key process for myelination which could explain the hypomyelinated state of patients. We also find that in both systems cells tend to form more neurites, likely losing their ability to differentiate, considering their progenitor state. We also report several alterations in the chromatin state of cells lacking Agc1, which confirms the hypothesis for which Agc1 is not a disease restricted only to metabolic alterations in the cells, but there is a profound shift of the regulatory state of these cells.

Decoding complex flow phenomena via Dynamic Mode Decomposition

In this thesis, the viability of the Dynamic Mode Decomposition (DMD) as a technique to analyze and model complex dynamic real-world systems is presented. This method derives, directly from data, computationally efficient reduced-order models (ROMs) which can replace too onerous or unavailable high-fidelity physics-based models. Optimizations and extensions to the standard implementation of the methodology are proposed, investigating diverse case studies related to the decoding of complex flow phenomena. The flexibility of this data-driven technique allows its application to high-fidelity fluid dynamics simulations, as well as time series of real systems observations. The resulting ROMs are tested against two tasks: (i) reduction of the storage requirements of high-fidelity simulations or observations; (ii) interpolation and extrapolation of missing data. The capabilities of DMD can also be exploited to alleviate the cost of onerous studies that require many simulations, such as uncertainty quantification analysis, especially when dealing with complex high-dimensional systems. In this context, a novel approach to address parameter variability issues when modeling systems with space and time-variant response is proposed. Specifically, DMD is merged with another model-reduction technique, namely the Polynomial Chaos Expansion, for uncertainty quantification purposes. Useful guidelines for DMD deployment result from the study, together with the demonstration of its potential to ease diagnosis and scenario analysis when complex flow processes are involved.

Adaptive wireless multimedia communication systems

In recent years, due to the rapid convergence of multimedia services, Internet and wireless communications, there has been a growing trend of heterogeneity (in terms of channel bandwidths, mobility levels of terminals, end-user quality-of-service (QoS) requirements) for emerging integrated wired/wireless networks. Moreover, in nowadays systems, a multitude of users coexists within the same network, each of them with his own QoS requirement and bandwidth availability. In this framework, embedded source coding allowing partial decoding at various resolution is an appealing technique for multimedia transmissions. This dissertation includes my PhD research, mainly devoted to the study of embedded multimedia bitstreams in heterogenous networks, developed at the University of Bologna, advised by Prof. O. Andrisano and Prof. A. Conti, and at the University of California, San Diego (UCSD), where I spent eighteen months as a visiting scholar, advised by Prof. L. B. Milstein and Prof. P. C. Cosman. In order to improve the multimedia transmission quality over wireless channels, joint source and channel coding optimization is investigated in a 2D time-frequency resource block for an OFDM system. We show that knowing the order of diversity in time and/or frequency domain can assist image (video) coding in selecting optimal channel code rates (source and channel code rates). Then, adaptive modulation techniques, aimed at maximizing the spectral efficiency, are investigated as another possible solution for improving multimedia transmissions. For both slow and fast adaptive modulations, the effects of imperfect channel estimation errors are evaluated, showing that the fast technique, optimal in ideal systems, might be outperformed by the slow adaptive modulation, when a real test case is considered. Finally, the effects of co-channel interference and approximated bit error probability (BEP) are evaluated in adaptive modulation techniques, providing new decision regions concepts, and showing how the widely used BEP approximations lead to a substantial loss in the overall performance.

Packet Level Coding for Mobile Broadcasting

The thesis deals with channel coding theory applied to upper layers in the protocol stack of a communication link and it is the outcome of four year research activity. A specific aspect of this activity has been the continuous interaction between the natural curiosity related to the academic blue-sky research and the system oriented design deriving from the collaboration with European industry in the framework of European funded research projects. In this dissertation, the classical channel coding techniques, that are traditionally applied at physical layer, find their application at upper layers where the encoding units (symbols) are packets of bits and not just single bits, thus explaining why such upper layer coding techniques are usually referred to as packet layer coding. The rationale behind the adoption of packet layer techniques is in that physical layer channel coding is a suitable countermeasure to cope with small-scale fading, while it is less efficient against large-scale fading. This is mainly due to the limitation of the time diversity inherent in the necessity of adopting a physical layer interleaver of a reasonable size so as to avoid increasing the modem complexity and the latency of all services. Packet layer techniques, thanks to the longer codeword duration (each codeword is composed of several packets of bits), have an intrinsic longer protection against long fading events. Furthermore, being they are implemented at upper layer, Packet layer techniques have the indisputable advantages of simpler implementations (very close to software implementation) and of a selective applicability to different services, thus enabling a better matching with the service requirements (e.g. latency constraints). Packet coding technique improvement has been largely recognized in the recent communication standards as a viable and efficient coding solution: Digital Video Broadcasting standards, like DVB-H, DVB-SH, and DVB-RCS mobile, and 3GPP standards (MBMS) employ packet coding techniques working at layers higher than the physical one. In this framework, the aim of the research work has been the study of the state-of-the-art coding techniques working at upper layer, the performance evaluation of these techniques in realistic propagation scenario, and the design of new coding schemes for upper layer applications. After a review of the most important packet layer codes, i.e. Reed Solomon, LDPC and Fountain codes, in the thesis focus our attention on the performance evaluation of ideal codes (i.e. Maximum Distance Separable codes) working at UL. In particular, we analyze the performance of UL-FEC techniques in Land Mobile Satellite channels. We derive an analytical framework which is a useful tool for system design allowing to foresee the performance of the upper layer decoder. We also analyze a system in which upper layer and physical layer codes work together, and we derive the optimal splitting of redundancy when a frequency non-selective slowly varying fading channel is taken into account. The whole analysis is supported and validated through computer simulation. In the last part of the dissertation, we propose LDPC Convolutional Codes (LDPCCC) as possible coding scheme for future UL-FEC application. Since one of the main drawbacks related to the adoption of packet layer codes is the large decoding latency, we introduce a latency-constrained decoder for LDPCCC (called windowed erasure decoder). We analyze the performance of the state-of-the-art LDPCCC when our decoder is adopted. Finally, we propose a design rule which allows to trade-off performance and latency.

Tecniche di ottimizzazione del software per sistemi su singolo chip per applicazioni di Nomadic Computing

I moderni sistemi embedded sono equipaggiati con risorse hardware che consentono l’esecuzione di applicazioni molto complesse come il decoding audio e video. La progettazione di simili sistemi deve soddisfare due esigenze opposte. Da un lato è necessario fornire un elevato potenziale computazionale, dall’altro bisogna rispettare dei vincoli stringenti riguardo il consumo di energia. Uno dei trend più diffusi per rispondere a queste esigenze opposte è quello di integrare su uno stesso chip un numero elevato di processori caratterizzati da un design semplificato e da bassi consumi. Tuttavia, per sfruttare effettivamente il potenziale computazionale offerto da una batteria di processoriè necessario rivisitare pesantemente le metodologie di sviluppo delle applicazioni. Con l’avvento dei sistemi multi-processore su singolo chip (MPSoC) il parallel programming si è diffuso largamente anche in ambito embedded. Tuttavia, i progressi nel campo della programmazione parallela non hanno mantenuto il passo con la capacità di integrare hardware parallelo su un singolo chip. Oltre all’introduzione di multipli processori, la necessità di ridurre i consumi degli MPSoC comporta altre soluzioni architetturali che hanno l’effetto diretto di complicare lo sviluppo delle applicazioni. Il design del sottosistema di memoria, in particolare, è un problema critico. Integrare sul chip dei banchi di memoria consente dei tempi d’accesso molto brevi e dei consumi molto contenuti. Sfortunatamente, la quantità di memoria on-chip che può essere integrata in un MPSoC è molto limitata. Per questo motivo è necessario aggiungere dei banchi di memoria off-chip, che hanno una capacità molto maggiore, come maggiori sono i consumi e i tempi d’accesso. La maggior parte degli MPSoC attualmente in commercio destina una parte del budget di area all’implementazione di memorie cache e/o scratchpad. Le scratchpad (SPM) sono spesso preferite alle cache nei sistemi MPSoC embedded, per motivi di maggiore predicibilità, minore occupazione d’area e – soprattutto – minori consumi. Per contro, mentre l’uso delle cache è completamente trasparente al programmatore, le SPM devono essere esplicitamente gestite dall’applicazione. Esporre l’organizzazione della gerarchia di memoria ll’applicazione consente di sfruttarne in maniera efficiente i vantaggi (ridotti tempi d’accesso e consumi). Per contro, per ottenere questi benefici è necessario scrivere le applicazioni in maniera tale che i dati vengano partizionati e allocati sulle varie memorie in maniera opportuna. L’onere di questo compito complesso ricade ovviamente sul programmatore. Questo scenario descrive bene l’esigenza di modelli di programmazione e strumenti di supporto che semplifichino lo sviluppo di applicazioni parallele. In questa tesi viene presentato un framework per lo sviluppo di software per MPSoC embedded basato su OpenMP. OpenMP è uno standard di fatto per la programmazione di multiprocessori con memoria shared, caratterizzato da un semplice approccio alla parallelizzazione tramite annotazioni (direttive per il compilatore). La sua interfaccia di programmazione consente di esprimere in maniera naturale e molto efficiente il parallelismo a livello di loop, molto diffuso tra le applicazioni embedded di tipo signal processing e multimedia. OpenMP costituisce un ottimo punto di partenza per la definizione di un modello di programmazione per MPSoC, soprattutto per la sua semplicità d’uso. D’altra parte, per sfruttare in maniera efficiente il potenziale computazionale di un MPSoC è necessario rivisitare profondamente l’implementazione del supporto OpenMP sia nel compilatore che nell’ambiente di supporto a runtime. Tutti i costrutti per gestire il parallelismo, la suddivisione del lavoro e la sincronizzazione inter-processore comportano un costo in termini di overhead che deve essere minimizzato per non comprometterre i vantaggi della parallelizzazione. Questo può essere ottenuto soltanto tramite una accurata analisi delle caratteristiche hardware e l’individuazione dei potenziali colli di bottiglia nell’architettura. Una implementazione del task management, della sincronizzazione a barriera e della condivisione dei dati che sfrutti efficientemente le risorse hardware consente di ottenere elevate performance e scalabilità. La condivisione dei dati, nel modello OpenMP, merita particolare attenzione. In un modello a memoria condivisa le strutture dati (array, matrici) accedute dal programma sono fisicamente allocate su una unica risorsa di memoria raggiungibile da tutti i processori. Al crescere del numero di processori in un sistema, l’accesso concorrente ad una singola risorsa di memoria costituisce un evidente collo di bottiglia. Per alleviare la pressione sulle memorie e sul sistema di connessione vengono da noi studiate e proposte delle tecniche di partizionamento delle strutture dati. Queste tecniche richiedono che una singola entità di tipo array venga trattata nel programma come l’insieme di tanti sotto-array, ciascuno dei quali può essere fisicamente allocato su una risorsa di memoria differente. Dal punto di vista del programma, indirizzare un array partizionato richiede che ad ogni accesso vengano eseguite delle istruzioni per ri-calcolare l’indirizzo fisico di destinazione. Questo è chiaramente un compito lungo, complesso e soggetto ad errori. Per questo motivo, le nostre tecniche di partizionamento sono state integrate nella l’interfaccia di programmazione di OpenMP, che è stata significativamente estesa. Specificamente, delle nuove direttive e clausole consentono al programmatore di annotare i dati di tipo array che si vuole partizionare e allocare in maniera distribuita sulla gerarchia di memoria. Sono stati inoltre sviluppati degli strumenti di supporto che consentono di raccogliere informazioni di profiling sul pattern di accesso agli array. Queste informazioni vengono sfruttate dal nostro compilatore per allocare le partizioni sulle varie risorse di memoria rispettando una relazione di affinità tra il task e i dati. Più precisamente, i passi di allocazione nel nostro compilatore assegnano una determinata partizione alla memoria scratchpad locale al processore che ospita il task che effettua il numero maggiore di accessi alla stessa.

Cooperative Wireless Networks for Localization

This thesis adresses the problem of localization, and analyzes its crucial aspects, within the context of cooperative WSNs. The three main issues discussed in the following are: network synchronization, position estimate and tracking. Time synchronization is a fundamental requirement for every network. In this context, a new approach based on the estimation theory is proposed to evaluate the ultimate performance limit in network time synchronization. In particular the lower bound on the variance of the average synchronization error in a fully connected network is derived by taking into account the statistical characterization of the Message Delivering Time (MDT) . Sensor network localization algorithms estimate the locations of sensors with initially unknown location information by using knowledge of the absolute positions of a few sensors and inter-sensor measurements such as distance and bearing measurements. Concerning this issue, i.e. the position estimate problem, two main contributions are given. The first is a new Semidefinite Programming (SDP) framework to analyze and solve the problem of flip-ambiguity that afflicts range-based network localization algorithms with incomplete ranging information. The occurrence of flip-ambiguous nodes and errors due to flip ambiguity is studied, then with this information a new SDP formulation of the localization problem is built. Finally a flip-ambiguity-robust network localization algorithm is derived and its performance is studied by Monte-Carlo simulations. The second contribution in the field of position estimate is about multihop networks. A multihop network is a network with a low degree of connectivity, in which couples of given any nodes, in order to communicate, they have to rely on one or more intermediate nodes (hops). Two new distance-based source localization algorithms, highly robust to distance overestimates, typically present in multihop networks, are presented and studied. The last point of this thesis discuss a new low-complexity tracking algorithm, inspired by the Fano’s sequential decoding algorithm for the position tracking of a user in a WLAN-based indoor localization system.

Mental states monitoring through passive Brain-Computer Interface systems

The monitoring of cognitive functions aims at gaining information about the current cognitive state of the user by decoding brain signals. In recent years, this approach allowed to acquire valuable information about the cognitive aspects regarding the interaction of humans with external world. From this consideration, researchers started to consider passive application of brain–computer interface (BCI) in order to provide a novel input modality for technical systems solely based on brain activity. The objective of this thesis is to demonstrate how the passive Brain Computer Interfaces (BCIs) applications can be used to assess the mental states of the users, in order to improve the human machine interaction. Two main studies has been proposed. The first one allows to investigate whatever the Event Related Potentials (ERPs) morphological variations can be used to predict the users’ mental states (e.g. attentional resources, mental workload) during different reactive BCI tasks (e.g. P300-based BCIs), and if these information can predict the subjects’ performance in performing the tasks. In the second study, a passive BCI system able to online estimate the mental workload of the user by relying on the combination of the EEG and the ECG biosignals has been proposed. The latter study has been performed by simulating an operative scenario, in which the occurrence of errors or lack of performance could have significant consequences. The results showed that the proposed system is able to estimate online the mental workload of the subjects discriminating three different difficulty level of the tasks ensuring a high reliability.

Modellbucher editi e inediti: dispositivi decodificatori della prassi teatrale brechtiana

I Modellbücher di Bertolt Brecht e del Berliner Ensemble si configurano come elementi dal carattere duplice e controverso, in bilico fra lo statuto di exempla per i futuri allestitori e quello di dispositivi interni al lavoro dell’ensemble, risultanti dinamiche della processualità scenica. L’interesse per la fotografia e il montaggio, che in Brecht assume molteplici forme – dando vita, ad esempio, ad opere come L’Abicí della guerra – si colloca al centro della composizione di tali strumenti, veri e propri esempi di assemblaggio di frammenti scenici. I Modellbücher trovano la propria ragion d’essere entro le dinamiche della regia brechtiana, di cui si prestano a tradurre le pratiche. La tesi indaga lo statuto dei Modellbücher, facendo riferimento alla storia degli studi e all’indagine archivistica condotta presso il Bertolt-Brecht-Archiv di Berlino fra 2018 e 2019.

La Bukowina e la "letteratura etnografica" di lingua tedesca

La Bukowina è una regione dell’impero asburgico ricordata come “eccezionale” per la compresenza e la tolleranza fra etnie, nonché per l’accoglienza unica della lingua tedesca. Questa narrazione proviene e dai nostalgici ricordi delle ultime generazioni di abitanti del luogo, che in seguito hanno vissuto la fine del “mondo di ieri”; e da un discorso propagandistico della Corona. La presente ricerca si propone di illustrare il paesaggio culturale della Bukowina nonché le reali modalità con cui la cultura dell’imperatore si insinuò nella regione. Nel decodificare il rapporto di superiorità con cui il tedesco si affacciava agli altri popoli – investigato con l’ausilio dei postcolonial studies –, il progetto propone come materia di studio un fenomeno letterario in voga a cavallo tra XIX e XX secolo, la “letteratura etnografica”, ispirato dall’affascinante molteplicità della Bukowina, nonché da una disciplina accademica che in quegli anni attirava molti studiosi e curiosi: l’etnografia. Questo tipo di letteratura si esprime attraverso diversi generi letterari, che spesso condividono il tratto della frammentarietà. Tra queste opere, la presente analisi si sofferma su tre esempi che mostrano un progressivo allontanamento dal metodo “scientifico” della disciplina etnografica, illustrando l’ampiezza dell’insieme “letteratura etnografica”: "Die Völkergruppen der Bukowina" (1884) di Ludwig Adolf Staufe Simiginowicz; "Aus Halb-Asien" (1876-1888) di Karl Emil Franzos; "Maghrebinische Geschichten" (1953), di Gregor von Rezzori.