The cross-situational word learning paradigm argues that word meanings can be approximated by word-object associations, computed from co-occurrence statistics between words and entities in the world. Lexicon acquisition involves simultaneously guessing (1) which objects are being talked about (the ”meaning”) and (2) which words relate to those objects. However, most modeling work focuses on acquiring meanings for isolated words, largely neglecting relationships between words or physical entities, which can play an important role in learning. Semantic parsing, on the other hand, aims to learn a mapping between entire utterances and compositional meaning representations where such relations are central. The focus is the mapping between meaning and words, while utterance meanings are treated as observed quantities. Here, we extend the joint inference problem of word learning to account for compositional meanings by incorporating a semantic parsing model for relating utterances to non-linguistic context. Integrating semantic parsing and word learning permits us to explore the impact of word-word and concept-concept relations. The result is a joint-inference problem inherited from the word learning setting where we must simultaneously learn utterance-level and individual word meanings, only now we also contend with the many possible relationships between concepts in the meaning and words in the sentence. To simplify design, we factorize the model into separate modules, one for each of the world, the meaning, and the words, and merge them into a single synchronous grammar for joint inference. There are three main contributions. First, we introduce a novel word learning model and accompanying semantic parser. Second, we produce a corpus which allows us to demonstrate the importance of structure in word learning. Finally, we also present a number of technical innovations required for implementing such a model.

The ratchet effect can be exploited in many types of research, yet few researchers pay attention to it. In this thesis, I investigate two examples of such effects in microfluidic devices, under the guidance of computational simulations. The first chapter provides a brief introduction to ratchet effects, electrophoresis, and swimming cells, topics directly related to the following chapters. The second chapter of this thesis studies the separation of charged spherical particles in various microfluidic devices. My work shows how to manipulate those particles with modified temporal asymmetric electric potentials. The rectification of randomly swimming bacteria in microfluidic devices has been extensively studied. However, there have been few attempts to optimize such rectification devices. Mapping such motion onto a lattice Monte Carlo model may suggest some new mathematical methods, which might be useful for optimizing the similar systems. Such a mapping process is introduced in chapter four.

L’effet du vieillissement sur les fonctions cognitives de haut niveau demeure encore relativement incompris. Cette recherche vise à mieux comprendre les différences interindividuelles de performances entre les individus jeunes et âgés par l’étude des processus de contrôle attentionnel mis en jeu dans les tâches de commutation attentionnelle. Dans unpremier temps, deux tâches d’alternance de type séries alternées ont été administrées à des adultes jeunes et âgés. Les résultats n’indiquent pas d’effet de l’âge sur les coûts d’alternance mesurés mais en revanche un coût d’alternance symétrique (étude 1) et des coûts d’alternance locaux et globaux différents selon les individus (étude 1 et 2). Dans un second temps, un travail de modélisation du fonctionnement cognitif à l’aide de l’architecture cognitive ACT-R a été réalisé. Il permet de tester la plausibilité des hypothèses de diminution de la vitesse de traitement (VT) et de diminution de la capacité de la mémoire de travail (CMT), à pouvoir reproduire les différences de performances entre jeunes et âgés. Lesrésultats des tests d’hypothèse pour les deux études réalisées indiquent que ces hypothèses ne reproduisent pas assez les effets empiriquement observés ce qui suppose que les hypothèses de diminution de la VT ou de la CMT sont insuffisantes pour expliquer les différences de performances individuelles observées. Cette étude met en évidence l’intérêt de la modélisation cognitive computationnelle dans la compréhension des processus sous-jacent le fonctionnement cognitif humain / The effect of aging on high level cognitive functions is still relatively misunderstood. The aim of this research is to better understand individual performance differences between young and elderly individuals, by studying the attentional control processes involved in task switching. Initially, two switching tasks were administered to young and elderly adults. The results indicate no age effect on the switching cost but show an asymmetric switching cost (Study 1) and differentlocal and global switching costs between individuals (study 1 and 2). In a second step, a computational cognitive modeling of cognitive functioning is built using the ACT-R architecture. It enables to test the veracity of the assumptions of reduced processing speed (VT) and reduced capacity of working memory (WCL), to reproduce the observed performance differences between young and old individuals. The results of the hypothesis tests for both studies indicate that these assumptions do not reproduce adequately the empirically observed effects, which implies that the assumptions of decreased VT or CMT are insufficient to explain the individual performance differences observed. This study highlights the interest of computational cognitive modeling in the understanding of the processes underlying human cognitive functioning

Similarity has been regarded as a primary means by which lexical representations are organized, and hence an important determinant of processing interactions between lexical items. A central question on lexical-semantics similarity is how it influences lexical processing. There have been much fewer investigations, however, on how lexical-semantic similarity might influence novel word learning. This dissertation work aimed to fill this gap by addressing one kind of lexical-semantic similarity, similarity among the novel words that are being learned concurrently (concurrent similarity), on the learning of phonological word forms. Importantly, it aimed to use tests that eliminated the real time processing confound at test so as to provide convincing evidence on whether learning was indeed affected by similarity. The first part of the dissertation addressed the effect of concurrent referent similarity on the learning of the phonological word forms. Experiment 1 used a naming test to provide evidence on the direction of the effect. Experiment 2 and Experiment 3 used the stem completion test and the recognition from mis-pronunciation test that controlled for real time processing between conditions. Then a 4-layer Hebbian Normalized Recurrent Network was also developed to provide even more convincing evidence on whether learning was affected (the connection weights). Consistently across the three tasks and the simulation, a detrimental effect of referent similarity on the phonological word form learning was revealed. The second part of the dissertation addressed the effect of cohort similarity on the learning of the phonological word forms. The recognition from mis-pronunciation on partial words was developed to control for real time processing between conditions so as to capture the effect of learning. We examined the effect of cohort similarity at different syllable positions and found a detrimental effect at the second syllable and non-effect at the third syllable. This is consistent with the previous finding that competition among cohorts diminishes as the stimulus is received, suggesting that the effect of cohort similarity depends on the status of competition dynamics among cohorts. The theoretical and methodological implications of this study are discussed.

abstract: The term Poly-Limb stems from the rare birth defect syndrome, called Polymelia. Although Poly-Limbs in nature have often been nonfunctional, humans have had the fascination of functional Poly-Limbs. Science fiction has led us to believe that having Poly-Limbs leads to augmented manipulation abilities and higher work efficiency. To bring this to life however, requires a synergistic combination between robot manipulation and wearable robotics. Where traditional robots feature precision and speed in constrained environments, the emerging field of soft robotics feature robots that are inherently compliant, lightweight, and cost effective. These features highlight the applicability of soft robotic systems to design personal, collaborative, and wearable systems such as the Soft Poly-Limb. This dissertation presents the design and development of three actuator classes, made from various soft materials, such as elastomers and fabrics. These materials are initially studied and characterized, leading to actuators capable of various motion capabilities, like bending, twisting, extending, and contracting. These actuators are modeled and optimized, using computational models, in order to achieve the desired articulation and payload capabilities. Using these soft actuators, modular integrated designs are created for functional tasks that require larger degrees of freedom. This work focuses on the development, modeling, and evaluation of these soft robot prototypes. In the first steps to understand whether humans have the capability of collaborating with a wearable Soft Poly-Limb, multiple versions of the Soft Poly-Limb are developed for assisting daily living tasks. The system is evaluated not only for performance, but also for safety, customizability, and modularity. Efforts were also made to monitor the position and orientation of the Soft Poly-Limbs components through embedded soft sensors and first steps were taken in developing self-powered compo-nents to bring the system out into the world. This work has pushed the boundaries of developing high powered-to-weight soft manipulators that can interact side-by-side with a human user and builds the foundation upon which researchers can investigate whether the brain can support additional limbs and whether these systems can truly allow users to augment their manipulation capabilities to improve their daily lives. / Dissertation/Thesis / Doctoral Dissertation Systems Engineering 2020

Over the past years, research efforts have focused on adding highly conductive nanoparticles, such as carbon nanotubes (CNTs) and graphene nanoplatelets (GNPs), into polymers to improve their electrical conductivity or to tailor their piezoresistive behavior. Resultant materials are typically described by the weight or volume fractions of their nanoparticles. The weight/volume fraction alone is a very global quantity, making it a poor evaluator of a doping configuration. Knowing which particles actually participate in improving electrical conductivity can optimize the doping strategy. Additionally, conductive particles are only capable of charge transfer over a very short range, thus most of them do not form part of the conduction path. Thus, understanding how these particles are arranged is necessary to increase their efficiency. First, this work focuses on polymers loaded with CNTs. A computational modeling strategy based on a full morphological analysis of the CNT network is presented to systematically analyze conductive networks and show how particles are arranged. A definition of loading efficiency is provided based on the results obtained from this morphology analysis. This study provides useful guidelines for designing these types of materials based on important features, such as representative volume element, nanotube tortuosity and length, tunneling cutoff distance, and efficiency. Second, a computational approach is followed to study the conductive network formed by hybrid particles in polymer nanocomposites. These hybrid particles are synthesized by growing CNTs on the surfaces of GNPs. The objective of this study is to show that the higher electrical conductivity of these composites is due to the hybrids forming a segregated structure. Polymers loaded with hybrid particles have shown a higher electrical conductivity compared with classical carbon fillers: only CNTs, only GNPs or mixed CNTs and GNPs. This is done to understand and compare the doping efficiency of the different types of nanoparticles. Finally, some parameters of the hybrid particle are studied: CNT density on GNPs, and CNT and GNP geometries. Recommendations to further improve the composite’s conductivity based on these parameters are presented. It is noted that this work is the first time the hybrid particle is studied through a computational approach.

Fabry disease is an inherited X-linked recessive disorder caused by mutations in the galactosidase alpha (GLA) gene, leading to deficiencies in α-galactosidase A (α-GAL) enzyme production. α-GAL, a lysosomal glycosidase, catalyzes the removal of a terminal α-galactose; however, loss of α-GAL activity leads to accumulation of globotriaosylceramide (an endogenous substrate) and the eventual onset of the disease. Approved treatments for Fabry disease include enzyme replacement therapy and pharmacological chaperone therapy. In the latter treatment, 1-deoxygalactonojirimycin (DGJ), a pharmacological chaperone, is administered to Fabry disease patients, leading to increased enzymatic activity. The DGJ iminosugar acts as a competitive inhibitor of α-GAL, and upon addition at sub-inhibitory concentrations, the α-GAL activity in the cell increases. At pH 7.5, the DGJ binds and stabilizes both wild type and mutant α-GAL and can thus drive the folding of the α-GAL protein (Guce 2011). DGJ has been clinically approved to treat a subset of the more than 900 known mutations in the GLA gene. These approvals come from the chaperone activity data published by Amicus Therapeutics (Benjamin 2017). However, these assays cost money, time, and effort to perform, and novel mutations are discovered annually. Using molecular dynamics energy calculations in the Schrödinger software package, we developed a model to predict successful chaperoning of the mutants. Overall, the results are directly applicable to Fabry disease, but could also be applied to the much larger family of protein folding diseases, including Alzheimer's, Parkinson's and Huntington's diseases.

Endothelial cells form the lining of the walls of blood vessels and are continuously subjected to mechanical stimuli from the blood flow. Microtubule-organizing center (MTOC),<br>also known as centrosome is a structure found in eukaryotic cells close to the nucleus. MTOC relocates relative to the nucleus when endothelial cells are exposed to shear stress which determines their polarization, thus it plays a critical role in cell migration and wound healing. The nuclear lamina, a mesh-like network that lies underneath the nuclear membrane, is composed of lamins, type V intermediate filament proteins. Mutations in LMNA gene that encodes A-type lamins cause the production of a mutant form of lamin A called progerin and leads to a rare premature aging disease known as Hutchinson-Gilford Progeria Syndrome<br><div>(HGPS). The goal of this study is to investigate how fluid flow affects the cytoskeleton of endothelial cells.</div><div><br></div>This thesis consists of two main sections; computational mechanical modeling and laboratory experimental work. The mechanical model was implemented using Ansys Workbench software as a tensegrity-based cellular model in order to simulate the state of an endothelial cell under the effects of induced shear stress from the blood fluid flow. This tensegrity-based cellular model - composed of a plasma membrane, cytoplasm, nucleus, microtubules, and<br><div>actin filaments - aims to understand the effects of the fluid flow on the mechanics of the cytoskeleton. In addition, the laboratory experiments conducted in this study examined the MTOC-nuclear orientation of endothelial cells under shear stress with the presence of wound healing. Wild-type lamin A and progerin-expressing BAECs were studied under static and sheared conditions.</div><div><br></div><div> Moreover, a custom MATLAB code was utilized to measure the MTOC-nuclear orientation</div>angle and classification. Results demonstrate that shear stress leads to different responses of the MTOC orientation between the wild-type and progerin-expressing cells around the vertical wound edge. Future directions for this study involve additional experimental work together with the improved simulation results to confirm the MTOC orientation<br>relative to the nucleus under shear stress.

Thesis advisor: Babak Momeni / Microbial communities are complex networks comprised of multiple species that are facilitating and inhibiting one another (as well as themselves). Currently, we lack an understanding of what mechanisms drive coexistence within these communities. We aimed to remedy this by studying the dynamics of coexisting communities, focusing on the complexity of their interaction networks, the impact of spatial dynamics, and the interplay of facilitating and inhibiting interactions. These limitations in our understanding prevent the furtherment of designing intentional communities for bioremediation, maintenance of healthy microbiota, and other functional communities. To better understand these microbial dynamics, we chose to address the problem from two fronts: computational modeling and exploring dynamics of cocultures. Through our 1-D model, spatial structure fostering more coexistence – especially when facilitation is present. For the coexistence assays, we determined that contact-dependent growth inhibition is a density dependent mechanism, and the use of a Tn-Seq mutant library to predict species interactions is possible, but needs further optimization to reconcile density dependent effects of interactions. / Thesis (MS) — Boston College, 2021. / Submitted to: Boston College. Graduate School of Arts and Sciences. / Discipline: Biology.

This diploma thesis focuses on computational modeling of the cell mechanical tests. The goal of this thesis is to build a cell model and to simulate compression test on this model. If necessary, the model should be adjusted so the model reflects real cell behavior. It was created the cell model reflecting cytoplasm, nucleus, membrane and cell cytoskeleton. Cytoskeleton was modeled as tensegrity structure. After this, the pressure test was simulated on this model. The behavior of the cell model and real cell was compared using the stress force. The stress force - cell deformation curve was markedly different for the cell model and the real cell. For this reason, the cytoplasm material model was adjusted. The difference between the curves was acceptable after this modification. It was found during computations that the cytoskeleton model influence on the cell load is minimal. These results does not reflects real cell behavior, which means that the model is considered inadequate for performing stress load simulation.