Exploring domain-informed and physics-guided learning in image-to-image translation

Image-to-image (i2i) translation networks can generate fake images beneficial for many applications in augmented reality, computer graphics, and robotics. However, they require large scale datasets and high contextual understanding to be trained correctly. In this thesis, we propose strategies for solving these problems, improving performances of i2i translation networks by using domain- or physics-related priors. The thesis is divided into two parts. In Part I, we exploit human abstraction capabilities to identify existing relationships in images, thus defining domains that can be leveraged to improve data usage efficiency. We use additional domain-related information to train networks on web-crawled data, hallucinate scenarios unseen during training, and perform few-shot learning. In Part II, we instead rely on physics priors. First, we combine realistic physics-based rendering with generative networks to boost outputs realism and controllability. Then, we exploit naive physical guidance to drive a manifold reorganization, which allowed generating continuous conditions such as timelapses.

Nanomechanical characterization of soft bioelectronic interfaces via modeling-informed atomic force microscopy

The field of bioelectronics involves the use of electrodes to exchange electrical signals with biological systems for diagnostic and therapeutic purposes in biomedical devices and healthcare applications. However, the mechanical compatibility of implantable devices with the human body has been a challenge, particularly with long-term implantation into target organs. Current rigid bioelectronics can trigger inflammatory responses and cause unstable device functions due to the mechanical mismatch with the surrounding soft tissue. Recent advances in flexible and stretchable electronics have shown promise in making bioelectronic interfaces more biocompatible. To fully achieve this goal, material science and engineering of soft electronic devices must be combined with quantitative characterization and modeling tools to understand the mechanical issues at the interface between electronic technology and biological tissue. Local mechanical characterization is crucial to understand the activation of failure mechanisms and optimizing the devices. Experimental techniques for testing mechanical properties at the nanoscale are emerging, and the Atomic Force Microscope (AFM) is a good candidate for in situ local mechanical characterization of soft bioelectronic interfaces. In this work, in situ experimental techniques with solely AFM supported by interpretive models for the characterization of planar and three-dimensional devices suitable for in vivo and in vitro biomedical experimentations are reported. The combination of the proposed models and experimental techniques provides access to the local mechanical properties of soft bioelectronic interfaces. The study investigates the nanomechanics of hard thin gold films on soft polymeric substrates (Poly(dimethylsiloxane) PDMS) and 3D inkjet-printed micropillars under different deformation states. The proposed characterization methods provide a rapid and precise determination of mechanical properties, thus giving the possibility to parametrize the microfabrication steps and investigate their impact on the final device.

Reti neurali "physics-informed" per modelli epidemiologici

La malattia COVID-19 associata alla sindrome respiratoria acuta grave da coronavirus 2 (SARS-CoV-2) ha rappresentato una grave minaccia per la salute pubblica e l’economia globale sin dalla sua scoperta in Cina, nel dicembre del 2019. Gli studiosi hanno effettuato numerosi studi ed in particolar modo l’applicazione di modelli epidemiologici costruiti a partire dai dati raccolti, ha permesso la previsione di diversi scenari sullo sviluppo della malattia, nel breve-medio termine. Gli obiettivi di questa tesi ruotano attorno a tre aspetti: i dati disponibili sulla malattia COVID-19, i modelli matematici compartimentali, con particolare riguardo al modello SEIJDHR che include le vaccinazioni, e l’utilizzo di reti neurali ”physics-informed” (PINNs), un nuovo approccio basato sul deep learning che mette insieme i primi due aspetti. I tre aspetti sono stati dapprima approfonditi singolarmente nei primi tre capitoli di questo lavoro e si sono poi applicate le PINNs al modello SEIJDHR. Infine, nel quarto capitolo vengono riportati frammenti rilevanti dei codici Python utilizzati e i risultati numerici ottenuti. In particolare vengono mostrati i grafici sulle previsioni nel breve-medio termine, ottenuti dando in input dati sul numero di positivi, ospedalizzati e deceduti giornalieri prima riguardanti la città di New York e poi l’Italia. Inoltre, nell’indagine della parte predittiva riguardante i dati italiani, si è individuato un punto critico legato alla funzione che modella la percentuale di ricoveri; sono stati quindi eseguiti numerosi esperimenti per il controllo di tali previsioni.

Meta-learning for informed model searches in Automated Machine Learning

Day by day, machine learning is changing our lives in ways we could not have imagined just 5 years ago. ML expertise is more and more requested and needed, though just a limited number of ML engineers are available on the job market, and their knowledge is always limited by an inherent characteristic of theirs: they are humans. This thesis explores the possibilities offered by meta-learning, a new field in ML that takes learning a level higher: models are trained on other models' training data, starting from features of the dataset they were trained on, inference times, obtained performances, to try to understand the relationship between a good model and the way it was obtained. The so-called metamodel was trained on data collected by OpenML, the largest ML metadata platform that's publicly available today. Datasets were analyzed to obtain meta-features that describe them, which were then tied to model performances in a regression task. The obtained metamodel predicts the expected performances of a given model type (e.g., a random forest) on a given ML task (e.g., classification on the UCI census dataset). This research was then integrated into a custom-made AutoML framework, to show how meta-learning is not an end in itself, but it can be used to further progress our ML research. Encoding ML engineering expertise in a model allows better, faster, and more impactful ML applications across the whole world, while reducing the cost that is inevitably tied to human engineers.

Up-regulation Of Dnam-1 And Nkp30, Associated With Improvement Of Nk Cells Activation After Long-term Culture Of Mononuclear Cells From Patients With Ovarian Neoplasia.

This study aimed at evaluating the functional activation and activating receptors expression on resting, short- and long-term NK and NK-like T cells from blood of ovarian neoplasia patients. Blood from patients with adnexal benign alterations (n = 10) and ovarian cancer (grade I-IV n = 14) were collected after signed consent. Effector cells activation was evaluated by the expression of the CD107a molecule. Short-term culture was conducted overnight with IL-2 and long-term culture for 21 days, by a method designed to expand CD56(+) lymphocytes. Short-term culture significantly increased NK cells activation compared to resting NK cells (p<0.05), however, the long-term procedure supported an even higher increase (p<0.001). Resting NK-like T cells showed poor activation, which was not altered by the culture procedures. The long-term culture effectively increased the expression of the activating receptors on NK and NK-like T cells, either by increasing the number of cells expressing a given receptor and/or by up-regulating their expression intensity. As a conclusion, the long-term culture system employed, resulted in a high number of functional NK cells. The culture system was particularly efficient on the up-regulation of NKp30 and DNAM-1 receptors on NK cells.

The Role Of Edaphic Environment And Climate In Structuring Phylogenetic Pattern In Seasonally Dry Tropical Plant Communities.

Seasonally dry tropical plant formations (SDTF) are likely to exhibit phylogenetic clustering owing to niche conservatism driven by a strong environmental filter (water stress), but heterogeneous edaphic environments and life histories may result in heterogeneity in degree of phylogenetic clustering. We investigated phylogenetic patterns across ecological gradients related to water availability (edaphic environment and climate) in the Caatinga, a SDTF in Brazil. Caatinga is characterized by semiarid climate and three distinct edaphic environments - sedimentary, crystalline, and inselberg -representing a decreasing gradient in soil water availability. We used two measures of phylogenetic diversity: Net Relatedness Index based on the entire phylogeny among species present in a site, reflecting long-term diversification; and Nearest Taxon Index based on the tips of the phylogeny, reflecting more recent diversification. We also evaluated woody species in contrast to herbaceous species. The main climatic variable influencing phylogenetic pattern was precipitation in the driest quarter, particularly for herbaceous species, suggesting that environmental filtering related to minimal periods of precipitation is an important driver of Caatinga biodiversity, as one might expect for a SDTF. Woody species tended to show phylogenetic clustering whereas herbaceous species tended towards phylogenetic overdispersion. We also found phylogenetic clustering in two edaphic environments (sedimentary and crystalline) in contrast to phylogenetic overdispersion in the third (inselberg). We conclude that while niche conservatism is evident in phylogenetic clustering in the Caatinga, this is not a universal pattern likely due to heterogeneity in the degree of realized environmental filtering across edaphic environments. Thus, SDTF, in spite of a strong shared environmental filter, are potentially heterogeneous in phylogenetic structuring. Our results support the need for scientifically informed conservation strategies in the Caatinga and other SDTF regions that have not previously been prioritized for conservation in order to take into account this heterogeneity.

Transtorno do déficit de atenção e hiperatividade: proposta de avaliação neuropsicológica para diagnóstico

Attention deficit hyperactivity disorder (ADHD) is characterized by a persistent pattern of inattention and/or hyperactivity/impulsivity. The aim of this research was to contribute more precisely to the diagnosis of ADHD, to propose a battery of neuropsychological assessment and to analyze the contribution of each test. We studied 10 matched pairs of children with ADHD and normal controls (7 to 11 years). Inclusion criteria were: presence of ADHD typical behavior, positive diagnosis of ADHD based on DSM-IV, normal IQ, normal neurological examination and parental consent. We used extensive neuropsychological battery. The results showed differential sensitivity for detection of attentional problems in children with ADHD, although most tests did not reach statistical significance. The item, errors, of WCST revealed statistically significant difference between the two groups: ADHD performance was inferior to controls . In conclusion the neuropsychological assessment battery used in this research contributed to the diagnosis of ADHD.

Programação esportiva na TV Brasil : a influência da ideologia e indústria cultural

Universidade Estadual de Campinas . Faculdade de Educação Física

O rugbi em cadeira de rodas : aspectos tecnicos e taticos e diretrizes para seu desenvolvimento

Universidade Estadual de Campinas . Faculdade de Educação Física

Influencia de diferentes informações quanto ao tempo de realização do exercicio sobre a atividade eletromiografica e consumo de oxigenio em ciclistas

Universidade Estadual de Campinas . Faculdade de Educação Física

Crescimento, composição corporal e performance motora em crianças e adolescentes de 7 a 14 anos provenientes de familias de baixo nivel socio-economico e participantes do projeto esporte solidario, São Luis - MA

Universidade Estadual de Campinas . Faculdade de Educação Física

Explorando relacões do grau de sedentarismo de alunos ingressantes na Universidade Estadual de Campinas

Universidade Estadual de Campinas. Faculdade de Educação Física

Atividades corporais : uma estrategia de adaptação biocultural numa comunidade rural do Amazonas

Universidade Estadual de Campinas . Faculdade de Educação Física

Corpo e loucura : historico das formas de intervenção sobre o corpo na psiquiatria

Universidade Estadual de Campinas . Faculdade de Educação Física

O papel do trabalho na transformação do homem em macaco : estudo sobre a disciplina no trabalho do jogador profissional do Sertãozinho F. C

Universidade Estadual de Campinas. Faculdade de Educação Física