9 research outputs found
Gestion énergétique intelligente pour la prévision d'énergie par l'apprentissage machine et l'intelligence artificielle
No full textRésumé
Ce projet se concentre sur l'analyse prédictive en utilisant différentes techniques de modélisation, notamment les modèles de base, les modèles hybrides, et la fusion tri-modèle. L'objectif est d'améliorer la précision des prévisions dans des contextes complexes, tels que l'analyse des séries chronologiques et d'autres données temporelles. Les modèles de base, tels que ARIMA, sont souvent utilisés pour des données avec des relations linéaires simples, mais présentent des limites pour des dynamiques plus complexes. Les modèles hybrides, comme SARIMA-SVR, combinent plusieurs approches pour capturer à la fois les relations linéaires et non linéaires. La fusion tri-modèle est une technique plus avancée qui combine trois types de modèles, comme SARIMA, SVM, et Random Forest, pour maximiser la robustesse et la précision des prévisions. Ce projet insiste également sur la préparation des données, en particulier la gestion des données manquantes, la normalisation et l'analyse des tendances. Cette combinaison de techniques avancées et d'une préparation rigoureuse des données vise à fournir des prévisions plus précises et à mieux comprendre les dynamiques sous-jacentes des données étudiées.
Abstract
This project focuses on predictive analytics using different modeling techniques, including baseline models, hybrid models, and tri-model fusion. The goal is to improve forecast accuracy in complex contexts, such as analyzing time series and other temporal data. Baseline models, such as ARIMA, are often used for data with simple linear relationships, but have limitations for more complex dynamics. Hybrid models, such as SARIMA-SVR, combine multiple approaches to capture both linear and nonlinear relationships. Tri-model fusion is a more advanced technique that combines three types of models, SARIMA, SVM, and Random Forest, to maximize forecast robustness and accuracy. This project also emphasizes data preparation, particularly handling missing data, normalization, and trend analysis. This combination of advanced techniques and rigorous data preparation aims to provide more accurate forecasts and better understand the underlying dynamics of the data studied
Reflections of MIT student entrepreneurs
No full textThis electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.Thesis: S.M. in Technology and Policy, Massachusetts Institute of Technology, School of Engineering, Institute for Data, Systems, and Society, 2019Cataloged from student-submitted PDF version of thesis. Vita.Includes bibliographical references (pages [79]-82).The study of social systems has increasingly relied on data collection and analysis to draw conclusions. In parallel, the research community has often tried to understand entrepreneurs quantitatively, e.g. by understanding which behaviours or personality traits most often correlated with entrepreneurial success. While a quantified representation is essential in modeling what is being studied, it hides away the mental processes that create behaviours. Arguably, the people who engage the most directly with systemic change are entrepreneurs. They have to create their own system (a company) and connect it to its wider network (clients, investors, etc.). Most importantly, the more uncommon their idea is, the less they can rely on existing frames to bring their ideas to fruition, and the more they have to reflect on the dynamics of their wider context and how their company can integrate to it.According to constructive-developmental theory (CDT), the ability to see context dynamically is not a personality trait, but a structure of perception. Our meaning-making--how we create that perception of the world--grows in complexity as we delve into our reflections. CDT makes the separation between a structure in which value is defined by the contexts that we are in, and the next more complex one, where the self defines value. This thesis is an early attempt at understanding the experience of systemic change, and the growth in perception that happens along-side it. By bringing CDT and associated theories of the mind into the experience of entrepreneurship, I sought to understand how entrepreneurs make meaning of themselves in the course of their growth, by asking them directly. This is the first known study that maps the meaning-making complexity of entrepreneurs using the subject-object interview, CDT's qualitative empirical research method.I found that, as perception became more intrinsic, the definition of value went from seeking to have what we wish to have, to leaving a legacy through impact, to acting in harmony with our meaning of value. Instead of seeking purpose, entrepreneurs who perceive intrinsically act on the one they already have. Instead of seeking value, they share it with others in all their interactions.by Othmane Benkirane.S.M. in Technology and PolicyS.M.inTechnologyandPolicy Massachusetts Institute of Technology, School of Engineering, Institute for Data, Systems, and Societ
Threat Modeling of Cyber-Physical Systems in Practice
No full textThis is a pre-print of Jamil, Ameerah-Muhsinah, Lotfi ben Othmane, and Altaz Valani. "Threat Modeling of Cyber-Physical Systems in Practice." arXiv preprint arXiv:2103.04226 (2021). Copyright 2021 The Author(s). Posted with permission.Traditional Cyber-physical Systems(CPSs) were not built with cybersecurity in mind. They operated on separate Operational Technology (OT) networks. As these systems now become more integrated with Information Technology (IT) networks based on IP, they expose vulnerabilities that can be exploited by the attackers through these IT networks. The attackers can control such systems and cause behavior that jeopardizes the performance and safety measures that were originally designed into the system. In this paper, we explore the approaches to identify threats to CPSs and ensure the quality of the created threat models. The study involves interviews with eleven security experts working in security consultation companies, software engineering companies, an Original Equipment Manufacturer (OEM),and ground and areal vehicles integrators. We found through these interviews that the practitioners use a combination of various threat modeling methods, approaches, and standards together when they perform threat modeling of given CPSs. key challenges practitioners face are: they cannot transfer the threat modeling knowledge that they acquire in a cyber-physical domain to other domains, threat models of modified systems are often not updated, and the reliance on mostly peer-evaluation and quality checklists to ensure the quality of threat models. The study warns about the difficulty to develop secure CPSs and calls for research on developing practical threat modeling methods for CPSs, techniques for continuous threat modeling, and techniques to ensure the quality of threat models
Regret Analysis of Learning-Based Linear Quadratic Gaussian Control with Additive Exploration
No full textThis thesis addresses the Learning-Based Control (LBC) of unknown partially observable systems in the Linear Quadratic (LQ) paradigm. In this setting of learning-based LQ control, the control action influences not only the control performance but also the rate at which the system is being learnt, causing a conflict between learning and control (exploration and exploitation), which is particularly challenging to address. This thesis aims to develop a novel LBC algorithm for unknown partially observable systems in the LQG setting that is computationally efficient and can guarantee an optimal exploration-exploitation trade-off, quantified by a metric called regret. The regret quantifies the cumulative performance gap between the LBC policy and the ideal controller having full knowledge of the true system dynamics. The contributions in this thesis involve a novel LBC algorithm deployed in a two-phase structure. The first phase involves injecting Gaussian input signals to obtain an initial system model. The subsequent second phase deploys the proposed LBC strategy in an episodic setting, where the model is updated for each episode, and the resulting updated LQG controller is applied with additive Gaussian signals for exploration. In addition, the thesis establishes strong theoretical guarantees on optimal regret growth.Mechanical Engineering | Systems and Contro
Modelling roughness effects on propagation of electromagnetic waves in a maritime environment: a hybrid approach
No full textWOSInternational audienceThis study presents work on radar links in the maritime environment. In order to understand and to optimise these links, the author have modelled electromagnetic wave propagation in the maritime environment, including in the presence of evaporation ducts. This modelling requires good propagation domain characterisation (maritime atmosphere, evaporation ducts etc.) and good modelling of the reflective effects of the sea surface (smooth or rough surface). In this study the author present a new method based on sea surface generation using sea spectra developed to model the sea surface roughness effects on electromagnetic wave propagation. Numerical results are given, in different configurations (duct height, transmitter altitude atmospheric conditions etc.), which enable the proposed model to be checked and highlight the interest of accurate roughness modelling in electromagnetic propagation over the sea surface
Finite-sample analysis of identification of switched linear systems with arbitrary or restricted switching
No full textFor the identification of switched systems with a measured switching signal,
this work aims to analyze the effect of switching strategies on the estimation
error. The data for identification is assumed to be collected from globally
asymptotically or marginally stable switched systems under switches that are
arbitrary or subject to an average dwell time constraint. Then the switched
system is estimated by the least-squares (LS) estimator. To capture the effect
of the parameters of the switching strategies on the LS estimation error,
finite-sample error bounds are developed in this work. The obtained error
bounds show that the estimation error is logarithmic of the switching
parameters when there are only stable modes; however, when there are unstable
modes, the estimation error bound can increase linearly as the switching
parameter changes. This suggests that in the presence of unstable modes, the
switching strategy should be properly designed to avoid the significant
increase of the estimation error
Homozygosity mapping of autosomal recessive demyelinating Charcot-Marie-Tooth neuropathy (CMT4H) to a novel locus on chromosome 12p11.21-q13.11
No full text[No abstract available]Azzedine H, 2003, AM J HUM GENET, V72, P1141, DOI 10.1086-375034; Baxter RV, 2002, NAT GENET, V30, P21, DOI 10.1038-ng796; Ben Othmane K, 1993, Hum Mol Genet, V2, P1625; Ben Othmane K, 1998, NEUROGENETICS, V2, P18, DOI 10.1007-s100480050047; Berciano J, 2003, CURR OPIN NEUROL, V16, P613, DOI 10.1097-01.wco.0000093105.34793.dd; Boerkoel CF, 2003, ANN NEUROL, V53, P400, DOI 10.1002-ana.10505; Bolino A, 2000, NAT GENET, V25, P17; Bolino A, 1996, HUM MOL GENET, V5, P1051, DOI 10.1093-hmg-5.7.1051; Boyle MET, 2001, NEURON, V30, P385, DOI 10.1016-S0896-6273(01)00296-3; De Sandre-Giovannoli A, 2002, AM J HUM GENET, V70, P726, DOI 10.1086-339274; Cuesta A, 2002, NAT GENET, V30, P22, DOI 10.1038-ng798; De Sandre-Giovannoli A, 2003, J Med Genet, V40, pe87, DOI 10.1136-jmg.40.7.e87; Delague V, 2000, AM J HUM GENET, V67, P236, DOI 10.1086-302980; Dib C, 1996, NATURE, V380, P152, DOI 10.1038-380152a0; Dyck PJ, 1993, PERIPHERAL NEUROPATH, P1094; ESCURAT M, 1990, J NEUROSCI, V10, P764; Guilbot A, 1999, EUR J HUM GENET, V7, P849, DOI 10.1038-sj.ejhg.5200382; Guillebaud J, 2001, PHARMACOEPIDEM DR S, V10, P21, DOI 10.1002-pds.560; Ikegami T, 1996, BIOCHEM BIOPH RES CO, V222, P107, DOI 10.1006-bbrc.1996.0705; Jordanova A, 2003, BRAIN, V126, P590, DOI 10.1093-brain-awg059; Kalaydjieva L, 1996, NAT GENET, V14, P214, DOI 10.1038-ng1096-214; Kalaydjieva L, 2000, AM J HUM GENET, V67, P47, DOI 10.1086-302978; Kennerson ML, 2001, AM J HUM GENET, V69, P883, DOI 10.1086-323743; Kuhlenbaumer G, 2002, J NEUROL, V249, P1629, DOI 10.1007-s00415-002-0946-3; LANDER ES, 1987, SCIENCE, V236, P1567, DOI 10.1126-science.2884728; LATHROP GM, 1985, AM J HUM GENET, V37, P482; LEFRANC G, 1975, ACTA ANTHROPOGENET, V1, P34; LeGuern E, 1996, HUM MOL GENET, V5, P1685, DOI 10.1093-hmg-5.10.1685; LUPSKI JR, 1991, CELL, V66, P219, DOI 10.1016-0092-8674(91)90613-4; Mersiyanova IV, 2000, AM J HUM GENET, V67, P37, DOI 10.1086-302962; Nelis E, 2004, J MED GENET, V41, P193, DOI 10.1136-jmg.2003.012633; Nelis E, 2002, NEUROLOGY, V59, P1865; Othmane Kamel Ben, 1999, Genomics, V62, P344, DOI 10.1006-geno.1999.6028; ROA BB, 1993, NAT GENET, V5, P269, DOI 10.1038-ng1193-269; ROA BB, 1993, NAT GENET, V5, P189, DOI 10.1038-ng1093-189; Rogers T, 2000, AM J HUM GENET, V67, P664, DOI 10.1086-303053; Senderek J, 2003, AM J HUM GENET, V73, P1106, DOI 10.1086-379525; SKRE H, 1974, CLIN GENET, V6, P98; Suter U, 2003, NAT REV NEUROSCI, V4, P714, DOI 10.1038-nrn1196; Timmerman V, 1999, NEUROLOGY, V52, P1827; Villanova M, 1998, NEUROMUSCULAR DISORD, V8, P392, DOI 10.1016-S0960-8966(98)00044-3; Warner LE, 1996, NEURON, V17, P451, DOI 10.1016-S0896-6273(00)80177-424252
Extracts of Ulva lactuca Induce Responses against Penicillium digitatum on Oranges
No full textPenicillium digitatum causes severe damage to postharvest and stored citrus fruits, making it necessary to seek acceptable solutions limiting environmental and sanitary impacts. The seaweed liquid extracts (SLEs) obtained from the green macroalgae Ulva lactuca have previously shown anti-fungal activities towards postharvest green mold due to P. digitatum. Moreover, treatment with aqueous and ethanolic extracts had an influence on disease incidence and lesion severity on wounded oranges “Valentia late”. Therefore, we studied the abilities of the U. lactuca extracts and chlorophyll a (as a standard) to activate defense mechanisms in oranges. The analyses showed modifications in the protein concentrations and activities of β-1,3-glucanase and peroxidase in orange tissues, with a maximum activity level reached at 72 h after treatments. It appears that the SLEs were able to fight P. digitatum infections in the absence of direct contact with the pathogen, reflecting that the induction of host resistance is also a probable mode of action. Ulva lactuca fractions are rich in chlorophyll a, polyphenols, organic acids and ulvans, which might be the key molecules involved in the defense mechanism of oranges’ albedo. The use of these natural substances could be an alternative way to control postharvest citrus rot
