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.

Predicting corticospinal excitability from pre-TMS EEG using convolutional neural networks

The amplitude of motor evoked potentials (MEPs) elicited by transcranial magnetic stimulation (TMS) of the primary motor cortex (M1) shows a large variability from trial to trial, although MEPs are evoked by the same repeated stimulus. A multitude of factors is believed to influence MEP amplitudes, such as cortical, spinal and motor excitability state. The goal of this work is to explore to which degree the variation in MEP amplitudes can be explained by the cortical state right before the stimulation. Specifically, we analyzed a dataset acquired on eleven healthy subjects comprising, for each subject, 840 single TMS pulses applied to the left M1 during acquisition of electroencephalography (EEG) and electromyography (EMG). An interpretable convolutional neural network, named SincEEGNet, was utilized to discriminate between low- and high-corticospinal excitability trials, defined according to the MEP amplitude, using in input the pre-TMS EEG. This data-driven approach enabled considering multiple brain locations and frequency bands without any a priori selection. Post-hoc interpretation techniques were adopted to enhance interpretation by identifying the more relevant EEG features for the classification. Results show that individualized classifiers successfully discriminated between low and high M1 excitability states in all participants. Outcomes of the interpretation methods suggest the importance of the electrodes situated over the TMS stimulation site, as well as the relevance of the temporal samples of the input EEG closer to the stimulation time. This novel decoding method allows causal investigation of the cortical excitability state, which may be relevant for personalizing and increasing the efficacy of therapeutic brain-state dependent brain stimulation (for example in patients affected by Parkinson’s disease).

Performance Analysis of Topological Quantum Error Correcting Codes

In the last few years there has been a great development of techniques like quantum computers and quantum communication systems, due to their huge potentialities and the growing number of applications. However, physical qubits experience a lot of nonidealities, like measurement errors and decoherence, that generate failures in the quantum computation. This work shows how it is possible to exploit concepts from classical information in order to realize quantum error-correcting codes, adding some redundancy qubits. In particular, the threshold theorem states that it is possible to lower the percentage of failures in the decoding at will, if the physical error rate is below a given accuracy threshold. The focus will be on codes belonging to the family of the topological codes, like toric, planar and XZZX surface codes. Firstly, they will be compared from a theoretical point of view, in order to show their advantages and disadvantages. The algorithms behind the minimum perfect matching decoder, the most popular for such codes, will be presented. The last section will be dedicated to the analysis of the performances of these topological codes with different error channel models, showing interesting results. In particular, while the error correction capability of surface codes decreases in presence of biased errors, XZZX codes own some intrinsic symmetries that allow them to improve their performances if one kind of error occurs more frequently than the others.

Convolutional Neural Network Architectures for Template Matching

The Neural Networks customized and tested in this thesis (WaldoNet, FlowNet and PatchNet) are a first exploration and approach to the Template Matching task. The possibilities of extension are therefore many and some are proposed below. During my thesis, I have analyzed the functioning of the classical algorithms and adapted with deep learning algorithms. The features extracted from both the template and the query images resemble the keypoints of the SIFT algorithm. Then, instead of similarity function or keypoints matching, WaldoNet and PatchNet use the convolutional layer to compare the features, while FlowNet uses the correlational layer. In addition, I have identified the major challenges of the Template Matching task (affine/non-affine transformations, intensity changes...) and solved them with a careful design of the dataset.

Machine learning and convolutional neural networks for melanomas identification

Il machine learning negli ultimi anni ha acquisito una crescente popolarità nell’ambito della ricerca scientifica e delle sue applicazioni. Lo scopo di questa tesi è stato quello di studiare il machine learning nei suoi aspetti generali e applicarlo a problemi di computer vision. La tesi ha affrontato le difficoltà del dover spiegare dal punto di vista teorico gli algoritmi alla base delle reti neurali convoluzionali e ha successivamente trattato due problemi concreti di riconoscimento immagini: il dataset MNIST (immagini di cifre scritte a mano) e un dataset che sarà chiamato ”MELANOMA dataset” (immagini di melanomi e nevi sani). Utilizzando le tecniche spiegate nella sezione teorica si sono riusciti ad ottenere risultati soddifacenti per entrambi i dataset ottenendo una precisione del 98% per il MNIST e del 76.8% per il MELANOMA dataset

Transmission Of Intra-cellular Genetic Information: A System Proposal.

One of the great challenges of the scientific community on theories of genetic information, genetic communication and genetic coding is to determine a mathematical structure related to DNA sequences. In this paper we propose a model of an intra-cellular transmission system of genetic information similar to a model of a power and bandwidth efficient digital communication system in order to identify a mathematical structure in DNA sequences where such sequences are biologically relevant. The model of a transmission system of genetic information is concerned with the identification, reproduction and mathematical classification of the nucleotide sequence of single stranded DNA by the genetic encoder. Hence, a genetic encoder is devised where labelings and cyclic codes are established. The establishment of the algebraic structure of the corresponding codes alphabets, mappings, labelings, primitive polynomials (p(x)) and code generator polynomials (g(x)) are quite important in characterizing error-correcting codes subclasses of G-linear codes. These latter codes are useful for the identification, reproduction and mathematical classification of DNA sequences. The characterization of this model may contribute to the development of a methodology that can be applied in mutational analysis and polymorphisms, production of new drugs and genetic improvement, among other things, resulting in the reduction of time and laboratory costs.

O papel da interação no processo de ensino-aprendizagem de português para alunos surdos em uma escola inclusiva

In this article, it is discussed the role of interaction in the process of teaching and learning Portuguese of deaf students at an inclusive school. In the context where the research took place, the hearing teacher does not understand sign language, and there are, in her classroom, hearing students and four deaf students, being three of them sign language users. As the communication between the hearing teacher and the deaf students occurred in different codes - Portuguese and Brazilian sign language - and having a social-interactional approach of language (MOITA LOPES, 1986; FREIRE, 1999), we observed if the interaction among the subjects enabled the deaf students to understand what was being taught. The results showed that the fact of having four deaf students in the same classroom allowed them to work in a cooperative way. Besides, the sign language became more visible in this institution. On the other hand, the interaction between the teacher and her deaf students revealed to be of little significance to the learning process of this small group.

Da formação e desenvolvimento profissional do professor de educação fisica a inovação educativa

Universidade Estadual de Campinas . Faculdade de Educação Física

Entre o fogo e o vento : as praticas de batuques e o controle das emoções

Universidade Estadual de Campinas . Faculdade de Educação Física

Oral health conditions among the elderly in Southeastern São Paulo State

OBJECTIVE: This study aimed to asses oral health conditions in a population aged 60 years and over living in Botucatu, Southeastern Brazil. A cross-sectional population-based study was carried out using a random sample (N=372) of the urban population aged 60 years and over from the city of Botucatu, in 2005. World Health Organization criteria and codes for oral health epidemiological surveys were used. Re-examination was carried out in 10% of individuals aiming to evaluate intra-examiner agreement. Statistical analysis was performed by one-way ANOVA or Kruskal-Wallis ANOVA, as applicable. Also, the t-test was used in the absence of homoscedasticity. Fisher's exact test was used for situations where the categories with less than five units were observed. Adjusted residuals and multiple-comparison analysis were conducted to identify associations between variable categories and subgroups. The intra-examiner agreement was 98% and Kappa statistics result was 0.95. Loss component represented 90.68% of DMF-T index, which was 29.85. The prevalence of edentulism was 63.17%. Upper and lower dentures were found in 80% and 58% respectively, with complete denture as the most commonly used. In those studied, 15% required upper and 38% lower dentures. There was more need for complete denture in both jaws. Approximately 20% had soft tissue alterations. For periodontal conditions, most sextants were excluded (81.81%). Periodontal pockets (4 - 5 mm) were seen in 11.29% of the examined individuals. The oral health status of the elderly population in Botucatu is poor, as well as in other Brazilian cities. The results of this study may help planning collective health actions, giving an accurate description of the oral problems among the elderly.

Uso da Classificação Internacional de Doenças na análise do absenteísmo odontológico

OBJETIVO: Comparar o uso das codificações da classificação de doenças e agravos em solicitações de afastamento do trabalho por motivo odontológico. MÉTODOS: Foram analisadas 240 solicitações emitidas em um serviço público federal entre janeiro de 2008 e dezembro de 2009. O uso da Classificação Estatística Internacional de Doenças e Problemas Relacionados à Saúde - Décima Revisão (CID-10) foi comparado ao sistema de Classificação Internacional de Doenças em Odontologia e Estomatologia (CID-OE). Foi determinada a especificidade da codificação nas solicitações de afastamento, bem como da codificação atribuída por peritos oficiais em inspeções indiretas, perícias e juntas odontológicas. RESULTADOS: Do total de atestados, 22,9% não apresentaram a CID, 7,1% apresentaram a CID-9, 3,3% a CID-OE e 66,7% a CID-10. A maioria das codificações foi concordante (55,1%), com maior especificidade nas codificações atribuídas após avaliação dos cirurgiões-dentistas peritos oficiais. CONCLUSÕES: É necessário aperfeiçoar a utilização da CID-10 entre os profissionais de Odontologia e perícia odontológica no trabalho. Sugere-se a incorporação do uso da CID-OE e da Classificação Internacional de Funcionamento, Incapacidade e Saúde para a análise dos afastamentos do trabalho, fornecendo dados relevantes para o monitoramento do absenteísmo por motivo odontológico.

Medidas de audição de pais de indivíduos com deficiência auditiva de herança autossômica recessiva

TEMA: avaliação audiológica de pais de indivíduos com perda auditiva de herança autossômica recessiva. OBJETIVO: estudar o perfil audiológico de pais de indivíduos com perda auditiva, de herança autossômica recessiva, inferida pela história familial ou por testes moleculares que detectaram mutação no gene GJB2, responsável por codificar a Conexina 26. MÉTODO: 36 indivíduos entre 30 e 60 anos foram avaliados e divididos em dois grupos: grupo controle, sem queixas auditivas e sem história familiar de deficiência auditiva, e grupo de estudos composto por pais heterozigotos em relação a genes de surdez de herança autossômica recessiva inespecífica ou portadores heterozigotos de mutação no gene da Conexina 26. Todos foram submetidos à audiometria tonal liminar (0,25kHz a 8), audiometria de altas freqüências (9kHz a 20) e emissões otoacústicas produtos de distorção (EOAPD). RESULTADOS: houve diferenças significativas na amplitude das EOAPD nas freqüências 1001 e 1501Hz entre os grupos, sendo maior a amplitude no grupo controle. Não houve diferença significativa entre os grupos para os limiares tonais de 0,25 a 20KHz. CONCLUSÃO: as EOAPD foram mais eficazes, em comparação com a audiometria tonal liminar, para detectar diferenças auditivas entre os grupos. Mais pesquisas são necessárias para verificar a confiabilidade destes dados.

Risk optimization of a steel frame communications tower subject to tornado winds

The structural engineering community in Brazil faces new challenges with the recent occurrence of high intensity tornados. Satellite surveillance data shows that the area covering the south-east of Brazil, Uruguay and some of Argentina is one of the world most tornado-prone areas, second only to the infamous tornado alley in central United States. The design of structures subject to tornado winds is a typical example of decision making in the presence of uncertainty. Structural design involves finding a good balance between the competing goals of safety and economy. This paper presents a methodology to find the optimum balance between these goals in the presence of uncertainty. In this paper, reliability-based risk optimization is used to find the optimal safety coefficient that minimizes the total expected cost of a steel frame communications tower, subject to extreme storm and tornado wind loads. The technique is not new, but it is applied to a practical problem of increasing interest to Brazilian structural engineers. The problem is formulated in the partial safety factor format used in current design codes, with all additional partial factor introduced to serve as optimization variable. The expected cost of failure (or risk) is defined as the product of a. limit state exceedance probability by a limit state exceedance cost. These costs include costs of repairing, rebuilding, and paying compensation for injury and loss of life. The total expected failure cost is the sum of individual expected costs over all failure modes. The steel frame communications, tower subject of this study has become very common in Brazil due to increasing mobile phone coverage. The study shows that optimum reliability is strongly dependent on the cost (or consequences) of failure. Since failure consequences depend oil actual tower location, it turn,,; out that different optimum designs should be used in different locations. Failure consequences are also different for the different parties involved in the design, construction and operation of the tower. Hence, it is important that risk is well understood by the parties involved, so that proper contracts call be made. The investigation shows that when non-structural terms dominate design costs (e.g, in residential or office buildings) it is not too costly to over-design; this observation is in agreement with the observed practice for non-optimized structural systems. In this situation, is much easier to loose money by under-design. When by under-design. When structural material cost is a significant part of design cost (e.g. concrete dam or bridge), one is likely to lose significantmoney by over-design. In this situation, a cost-risk-benefit optimization analysis is highly recommended. Finally, the study also shows that under time-varying loads like tornados, the optimum reliability is strongly dependent on the selected design life.

Age and metallicity of star clusters in the Small Magellanic Cloud from integrated spectroscopy

Context. Analysis of ages and metallicities of star clusters in the Magellanic Clouds provide information for studies on the chemical evolution of the Clouds and other dwarf irregular galaxies. Aims. The aim is to derive ages and metallicities from integrated spectra of 14 star clusters in the Small Magellanic Cloud, including a few intermediate/old age star clusters. Methods. Making use of a full-spectrum fitting technique, we compared the integrated spectra of the sample clusters to three different sets of single stellar population models, using two fitting codes available in the literature. Results. We derive the ages and metallicities of 9 intermediate/old age clusters, some of them previously unstudied, and 5 young clusters. Conclusions. We point out the interest of the newly identified as intermediate/old age clusters HW1, NGC 152, Lindsay 3, Lindsay 11, and Lindsay 113. We also confirm the old ages of NGC 361, NGC 419, Kron 3, and of the very well-known oldest SMC cluster, NGC 121.

PARTITIONED ANALYSIS FOR DIMENSIONALLY-HETEROGENEOUS HYDRAULIC NETWORKS

In this work an iterative strategy is developed to tackle the problem of coupling dimensionally-heterogeneous models in the context of fluid mechanics. The procedure proposed here makes use of a reinterpretation of the original problem as a nonlinear interface problem for which classical nonlinear solvers can be applied. Strong coupling of the partitions is achieved while dealing with different codes for each partition, each code in black-box mode. The main application for which this procedure is envisaged arises when modeling hydraulic networks in which complex and simple subsystems are treated using detailed and simplified models, correspondingly. The potentialities and the performance of the strategy are assessed through several examples involving transient flows and complex network configurations.