174 research outputs found

    Ocular manifestations of rheumatoid arthritis and their correlation with anti-cyclic citrullinated peptide antibodies

    No full text
    Ammapati Paul Pandian Vignesh, Renuka Srinivasan Department of Ophthalmology, Jawaharlal Institute of Postgraduate Medical Education and Research, Pondicherry, India Purpose: To study the ocular manifestations of rheumatoid arthritis and to correlate the role of anti-cyclic citrullinated peptide antibody (anti-CCP antibody) with the ocular manifestations.Methods: Three-hundred and ninety-two eyes of the 196 rheumatoid arthritis patients who attended the ophthalmology outpatient department underwent a detailed ocular examination using slit lamp biomicroscopy and ophthalmoscopy. The tear function of all the patients was assessed using Schirmer’s test, tear film break-up time and ocular surface staining. The anti-CCP antibody titers for all the rheumatoid arthritis patients were estimated using enzyme-linked immunosorbent assay tests.Results: Seventy-seven patients (135 eyes, 39%) out of the 196 patients studied had ocular manifestations typical of rheumatoid arthritis. Dry eye was the most common manifestation (28%, 54 patients). Of the patients, 78% was females (60 patients). The mean duration of rheumatoid arthritis in patients with ocular manifestations was 5.4±2.7 years and without ocular manifestations was 2.1±1.6years. Three percent of the patients had episcleritis (six patients). Scleritis was present in 2% of the patients (four patients). Peripheral ulcerative keratitis and sclerosing keratitis was present in 1% of the population each (two patients each). Eighty-five percent (66 patients) had bilateral manifestations 15% (eleven patients) had unilateral manifestations. There was a strong association between the presence of anti-CCP antibodies and ocular manifestations of rheumatoid arthritis which was shown by the statistically significant P-value of <0.0001.Conclusion: Ocular manifestations are a significant part of the extra-articular manifestation of rheumatoid arthritis. Dry eye was the most common ocular manifestation. There was a statistically significant association between the presence of anti-CCP antibodies specific to rheumatoid arthritis and the ocular manifestations. Keywords: rheumatoid arthritis, ocular manifestations, anti-CCP antibodies, dry eye, scleritis, peripheral ulcerative keratiti

    Blockchain Social Theory and Societal Impact

    No full text
    We face a daily reality such that very interesting outlook changes are testing the fundamental organizations that trust is based on. The absence of trust related with information administration, which is frequently knowledgeable about the type of information breaks or essentially an adaptation of our information without our consent as well as impetus to take part in this arising decentralization of designs, is basically difficult centralization as states, monetary establishments, ventures, and associations. We perceive this trust gap posing a threat to the very institutions on which we have relied, such as financial institutions, private businesses, and government agencies. As people keep on advancing into more decentralized and self- administering (or semi-independent) associations, a new “common agreement” is fundamental. Although there has been significant discussion regarding blockchain applications and prospective results in the FinTech industry, little has been done to examine how client-driven blockchain advancements can enable a variety of purposes outside of banking. This article plans to add to that assortment of information by inspecting blockchain innovation’s likely applications, as well as its restrictions, in regions where social effect crosses, like basic liberties. This likewise examines whether blockchain innovation and its center functional standards - like decentralization, straightforwardness, equity, and responsibility - can assist with restricting unnecessary internet-based observation, oversight, and common liberties infringement that are worked with by the developing dependence on a couple of elements to control admittance to data on the web. With regards to the expected effect of blockchain innovation on society, what is conceivable and what ought to be kept away from. With the proviso that an administration instrument should be indicated, we will see further improvement in computerized power. The data introduced in this position paper upholds the possibility that blockchain and individual tokenization could make another common agreement

    2010 Seventh International Conference on Information Technology: New Generations

    No full text
    Isaac Macwan (with Hassan Bajwa, Vignesh Veerapandian, and Xinghao Chen) is a contributing author, VHDL Implementation of High-Performance and Dynamically Configures Multi-port Cache Memory, pp. 1212-1216.https://digitalcommons.fairfield.edu/engineering-books/1057/thumbnail.jp

    How the planning, engineering and politics of transportation established, preserves and perpetuates the automobile city

    No full text
    Thesis (M.C.P.)--Massachusetts Institute of Technology, Dept. of Urban Studies and Planning; and, (S.M. in Transportation)--Massachusetts Institute of Technology, Dept. of Civil and Environmental Engineering, 2012.This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.Cataloged from student submitted PDF version of thesis. Page [167] blank.Includes bibliographical references (p. 161-166).The last eight decades of urban transportation planning and engineering in the United States have been dominated by the hegemony of the automobile. Auto-oriented planning of the transportation and land use system has had a profound impact on the built environment both in greenfield developments and neighborhoods that predated the auto. The pedestrian quality of cities has been eroded by the automobile, and urban renewal in the United States erased many neighborhoods strongly oriented around walking and transit use. Equally pervasive as the auto itself is the place for the car in the institutional cultures and practices involved in shaping the city. The shortcomings of mobility-oriented transportation planning have been well critiqued, even from the very early days of Interstate building. In recent decades there has been a flurry of interest in articulating sustainable transportation policies to provide multi-modal accessibility and to consider the interactions between transportation, land use, and other policy realms such as health, energy, environment and equity. The current impending crisis of aging and ailing highway structures in the United States presents a momentous opportunity to reassess the need and purpose of such infrastructure, and to rebuild, reconceptualize, or remove it in a matter more consistent with current policy goals and planning processes - rather than the ones in place when initially built. Despite the interest, need and opportunity to reconceptualize aging infrastructure in America to support a more sustainable reshaping of land use and activity patterns, the potential to do so is heavily impaired by a transportation planning process that is still dominated by the tools, methods and assumptions, political biases, procedural failures, and instilled human behaviors of the first highway-building era. The McGrath Highway in Somerville, MA is used as a case study to discuss how persistence of 1950s technical, procedural and political dysfunctions threaten to undermine this opportunity. Short-term actions and strategies to avoid this impending fate are suggested for McGrath Highway with applicability to a wider national context of similar opportunities.by Vignesh Krishnamurthy.S.M.in TransportationM.C.P

    Novel roles of the protein deubiquitinase USP14 in proteostasis regulation and cell stress responses

    No full text
    Ubiquitin specific protease-14 (USP14) is a deubiquitinating enzyme that adaptively regulates proteostasis to determine the fate of cellular processes and their characteristics. USP14 has been implicated in the progression of cancer, and neurodegeneration, as well as in metabolic disorders such as fatty liver and diabetes. Small-molecule compounds inhibiting USP14 were shown to have beneficial effects in reducing tumor growth, progression, and metastasis. At the molecular and structural level, the importance of USP14 in regulating proteasome-mediated degradation has been elucidated in detail. However, the context-dependent functions of USP14 in neuronal and non-neuronal cells, and its importance in context of neurological disorders such as Parkinson’s disease (PD) has not been characterized extensively. In this thesis, I have studied the role of USP14 in proteasome-autophagy network and cell stress responses such as oxidative stress utilizing a comprehensive set of biochemical and molecular biology methods in neuronal cell lines, primary cortical neurons, and follicular thyroid cancer cells. In addition, I have identified a molecular signaling including protein phosphorylation that influences USP14 and its interaction with the 26S proteasomes. In study I, the results show that USP14 interacts with the chaperone protein heat shock cognate 71 kDa protein (HSC70), as identified with quantitative interactomics mass spectrometry. Interaction of USP14 with HSC70 increased upon proteasome inhibition or N-terminal ubiquitin-like domain (UBL) domain mutation (W58A). On the other hand, W58A mutation of USP14 abrogated its interaction with the 19S regulatory particle (19S RP) subunit, proteasome 26S subunit ubiquitin receptor, non-ATPase 2 (PSMD2) in the proteasome. We further assessed the role of USP14-HSC70 interaction in the context of Huntington’s disease (HD) utilizing a striatal neuronal cell-line expressing mutant 109 polyglutamine Huntingtin (Htt) repeats (109Q) compared to control striatal neurons expressing 7 polyglutamine Htt repeats (7Q). Striatal neuronal cells expressing mutant 109Q Htt exhibited reduced USP14 levels compare to the control. Furthermore, expression of the W58A-USP14 in striatal neuronal cells promoted the presence of gamma-aminobutyric acid receptor-associated protein (GABARAP)-positive autophagosomes in an HSC70-dependent manner. These findings elucidated that USP14 upon dissociation from the proteasome, could be functionally important in regulating HD-related autophagy in a HSC70-dependent manner. To elucidate further the importance of USP14 in PD-related neuronal signaling, we generated a USP14-deleted SH-SY5Y human dopaminergic-like neuronal cell line. Clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 gene deletion of USP14 in SH-SY5Y cells revealed compromised 26S proteasomes, 20S core particle (CP), and increased levels of phosphorylated at serine129 α-synuclein (pS129 α-synuclein), α-synuclein. Furthermore, USP14-deleted cells showed elongated mitochondria, and elevated levels of reactive oxygen species (ROS). The increased levels of α-synuclein and pS129 α-synuclein in USP14-deleted cells were ameliorated by using the antioxidant, N-acetyl cysteine amide (NACA). These suggest that compromised proteasome and elevated ROS arising from USP14 deletion could be one of the important contributors to the levels of α-synuclein and pS129 α-synuclein in these cells. Using proteomics, we discovered that USP14 is phosphorylated at Serine 143 (pS143) in SH-SY5Y cells, which affected the binding of USP14 to the 26S proteasomes, as shown by molecular modeling and functional assays. Functionally, expression of pS143-USP14 mimetic in USP14-deleted SH-SY5Y cells was found to counteract ROS, pS129 α-synuclein, and α-synuclein levels. In view of these findings, USP14 could be an important target to ameliorate proteostasis defects and oxidative stress observed in PD. Using a model for thyroid cancer, we observed that the levels of USP14 were decreased in ML1 human follicular thyroid cancer cells. Treatment of ML1 cells with IU1 elevated the inhibited microtubule associated protein 1 light chain 3 beta (LC3B)-associated autophagy flux. In contrast, IU1 inhibited LC3B-associated autophagy flux in control primary thyroid cells. As measured by 3-(4, 5-dimethylthiazolyl-2)-2, 5-diphenyltetrazolium bromide (MTT) assay and radioactive thymidine incorporation assays, IU1 also reduced the proliferation of ML1 cells while this effect was prominent only at higher doses in primary thyroid cells. These findings show that USP14 can be a possible molecular target for treatment against follicular thyroid cancer cells as also shown for other cancer types. IU1 may elicit a cell-type specific response at a molecular level, which is important for future consideration. The findings presented in the thesis increase our understanding of USP14 and pS143-USP14 in regulating the abnormal 26S proteasome function, autophagy, and oxidative stress that are often associated with neurodegenerative diseases such as PD and HD. Inhibiting USP14 with IU1 could be a viable strategy in thyroid cancer models. USP14 may play an important role in increasing GABARAP-associated neuronal autophagy in HD, promoting α-synuclein clearance and counteracting oxidative stress that are relevant in the context of PD. Therefore, inhibiting USP14 with small-molecular compounds such as IU1 may not be beneficial in all disease pathologies. Importantly, the identified USP14 phosphorylation at S143 could largely affect its ubiquitin-like domain (UBL)-domain conformation and interactions with the proteasomes. This could be critical in regulating the downstream stress responses in different cell types. Further examination of these molecular aspects may aid in identifying targeted strategies to fine-tune proteostasis and cellular physiology.Ubikitiinispesifinen proteaasi 14 (USP14) on deubikitinoiva entsyymi, joka säätelee adaptiivisesti proteostaasia määrittääkseen soluprosessien kohtalon ja niiden ominaisuudet. USP14 on osallisena syövän ja neurodegeneratiivisten sairauksien etenemisessä sekä metabolisissa häiriöissä, kuten rasvamaksassa ja diabeteksessä. Pienimolekyylisillä USP14:ää inhiboivilla yhdisteillä osoitettiin olevan hyödyllisiä vaikutuksia kasvaimen kasvun, etenemisen ja etäpesäkkeiden vähentämisessä. Molekyyli- ja rakenteellisella tasolla USP14:n merkitys proteasomin välittämän hajotuksen säätelyssä on selvitetty yksityiskohtaisesti. USP14:n kontekstiriippuvaisia toimintoja hermo- ja muissa soluissa sekä sen merkitystä neurologisten sairauksien, kuten Parkinsonin taudin (PD), yhteydessä ei ole kuitenkaan kuvattu laajasti. Tässä väitöskirjassa olen tutkinut USP14:n roolia proteasomi-autofagiaverkostossa ja solustressivasteissa, kuten oksidatiivisessa stressissä, hyödyntäen laajaa valikoimaa biokemiallisia ja molekyylibiologisia menetelmiä hermosolulinjoissa, primaarisissa kortikaalisissa hermosoluissa ja follikulaarisissa kilpirauhassyöpäsoluissa. Lisäksi olen tunnistanut molekyylisignaloinnin, joka pitää sisällään proteiinifosforylaation, joka vaikuttaa USP14:ään ja sen vuorovaikutukseen 26S-proteasomien kanssa. Ensimmäisessä tutkimuksessa tulokset osoittavat, että USP14 vuorovaikuttaa kaitsijaproteiini HSC70 (heat shock cognate 71 kDa) kanssa, kuten havaittiin kvantitatiivisella interaktomiikkamassaspektrometrialla. USP14:n vuorovaikutus HSC70:n kanssa lisääntyi proteasomin eston tai N-terminaalisen ubikitiinin kaltaisen (UBL) domeenin mutaation (W58A) jälkeen. Toisaalta USP14 W58A-mutaatiolla menetettiin sen vuorovaikutus 19S säätelypartikkelin (19S RP) alayksikön, proteasomi 26S alayksikön ubikitiinireseptorin, ei-ATPaasin 2 (PSMD2), kanssa proteasomissa. Arvioimme edelleen USP14-HSC70-vuorovaikutuksen roolia Huntingtonin taudin (HD) yhteydessä käyttäen striataalista hermosolulinjaa, joka ilmentää mutantteja 109-polyglutamiini Huntingtin (Htt) toistoja (109Q) verrattuna kontrolleina toimineisiin striataalisiin hermosoluihin, jotka ilmentävät 7-polyglutamiini Htt –toistoja (7Q). 109Q Htt –mutaatiota ilmentävien striataalisten hermosolujen USP14-tasot olivat laskeneet verrattuna kontrolliin. Sen lisäksi striataalisten hermosolujen W58A-USP14-ekspressio lisäsi gamma-aminovoihapporeseptoriin liittyvän proteiinin (GABARAP)-positiivisten autofagosomien määrää HSC70-riippuvaisella tavalla. Nämä havainnot selvensivät, että proteasomista irtoavalla USP14:llä voisi olla toiminnallisesti tärkeä rooli HD:hen liittyvän autofagian säätelyssä HSC70-riippuvaisella tavalla. USP14:n PD:hen liittyvän neuronisignaloinnin tärkeyden selventämiseksi loimme ihmisen dopaminergisen kaltaisen SH-SY5Y-hermosolulinjan, jolta poistettiin USP14. CRISPR/Cas9-geenin (Clustered regularly interspaced short palindromic repeats) poisto SH-SY5Y-soluissa paljasti heikentyneet 26S-proteasomit, 20S-ydinpartikkelin (CP) ja kohonneet seriinissä 129 fosforyloidun α-synukleiinin (pS129 α-synukleiini) ja α-synukleiinin tasot. Lisäksi USP14-poistettujen solujen mitokondriot olivat pidentyneet ja reaktiivisen happilajin (ROS) tasot kohonneet. Kohonneita USP14-poistettujen solujen α-synukleiinin ja pS129 α-synukleiinin tasoja nostettiin edelleen käyttämällä antioksidanttia, N-asetyylikysteiiniamidia (NACA). Tämä viittaa siihen, että USP14:n poiston aiheuttama heikentynyt proteasomi ja kohonnut ROS voivat olla yksi tärkeä tekijä α-synukleiinin ja pS129 α-synukleiinin tasojen nousussa näissä soluissa. Proteomiikan avulla havaitsimme, että USP14 fosforyloituu seriinin 143 (pS143) kohdalla SH-SY5Y-soluissa, mikä vaikutti USP14:n sitoutumiseen 26S-proteasomeihin, kuten molekyylimallinnus ja toiminnalliset kokeet osoittivat. Toiminnallisesti pS143-USP14:n jäjennöksen ilmentämisen USP14-poistetuissa SH-SY5Y-soluissa havaittiin vastustavan ROS:ää, pS129 α-synukleiinia ja α-synukleiinin tasoja. Näiden havaintojen perusteella USP14 voisi olla tärkeä kohde PD:ssä havaittujen proteostaasin häiriöiden kohentamiseksi ja oksidatiivisen stressin lievittämiseksi. Kilpirauhassyövän mallia käyttäen havaitsimme, että USP14-tasot olivat alentuneet ihmisen follikulaarisissa ML1-kilpirauhassyöpäsoluissa. ML1-solujen hoito IU1:llä nosti estynyttä mikrotubulukseen liittyvän proteiinin 1 kevytketju 3-beeta (LC3B) -liitännäisen autofagian tahtia. Sitä vastoin IU1 hidasti LC3B-liitännäisen autofagian tahtia kontrollisoluina toimineissa primaarisissa kilpirauhassoluissa. Kuten 3-(4, 5-dimetyylitiasolyyli-2)-2, 5-difenyyli-tetrazoliumbromidi (MTT) -testillä ja radioaktiivisella tymidiinin inkorporaatiotesteillä mitattiin, IU1 vähensi ML1-solujen jakaantumista, kun taas tämä vaikutus oli merkittävä vain suuremmilla annoksilla primaarisissa kilpirauhassoluissa. Nämä havainnot osoittavat, että USP14 voi olla mahdollinen molekyylikohde follikulaaristen kilpirauhassyöpäsolujen hoidossa, kuten myös muissa syöpätyypeissä on osoitettu. IU1 voi saada aikaan solutyyppispesifisen vasteen molekyylitasolla, mikä on tärkeää tulevan pohdinnan kannalta. Väitöskirjassa esitetyt havainnot lisäävät ymmärrystämme USP14:stä ja pS143-USP14:stä 26S-proteasomien toimintahäiriöiden, autofagian ja oksidatiivisen stressin säätelyssä, jotka liittyvät usein neurodegeneratiivisiin sairauksiin, kuten PD:hen ja HD:hen. USP14:n estäminen IU1:llä voisi olla toimiva strategia kilpirauhassyöpämalleissa. USP14 saattaa olla tärkeässä roolissa lisäämällä GABARAP-liitännäistä hermosolujen autofagiaa HD:ssa, edistäen α-synukleiinin poistumista ja vastustaen oksidatiivista stressiä, jotka ovat olennaisia PD:n kannalta. Siksi USP14:n estäminen pienimolekyylisillä yhdisteillä, kuten IU1:llä, ei välttämättä ole suotuisaa kaikissa tautitiloissa. Merkittävää on, että havaittu USP14:n fosforylaatio S143:n kohdalla voi vaikuttaa suuresti sen ubikitiinin kaltaisen domeenin (UBL)-domeenin konformaatioon ja vuorovaikutuksiin proteasomien kanssa. Tämä voisi olla ratkaisevaa solutyyppikohtaisten stressireaktioiden säätelyssä. Näiden molekyylitason havaintojen tarkempi tutkiminen voi auttaa tunnistamaan kohdennettuja strategioita proteostaasin ja solufysiologian hienosäätöön.ei saavutettav

    Robustheit von tiefen neuronalen Netzen gegenüber adversen, randständigen und verzerrten Beispielen

    No full text
    Recently Deep Neural Network (DNN) models have shown remarkable successes on several tasks including classification, domain translation etc. However, those methods typically do not perform well on samples lying on relatively low-density areas of the data distribution, where the model was not well trained. In this thesis, we analyze the effect of different types of noise on the predictions of different DNN-based applications. In particular, for classification based models, we propose a generalized framework for crafting adversarial examples in a blackbox attack setting. As defense against such adversarial examples, we propose a novel algorithm called MALADE, which drives the given off-manifold input towards the high density regions of the data generating distribution with intrinsic knowledge of the perceptual decision boundary during inference. For domain translation based models, we propose to drive the unsuccessful fringe examples towards the data manifold by cooling the input test distribution using Langevin dynamics. We demonstrate qualitatively and quantitatively that our strategy enhances the robustness of state-of-the-art methods for classification as well as for domain translation tasks. Taking medical imaging as an exemplar use-case of DNN-based classification, we evaluate the robustness of pretraining and self-supervision strategies to input distortions and bias.In letzter Zeit haben Tiefe Neuronale Netze Modelle bemerkenswerte Erfolge bei verschiedenen Aufgaben wie Klassifizierung, Domänenübersetzung usw. gezeigt. Allerdings liefern diese Methoden in der Regel weniger gute Ergebnisse bei Stichproben, die in Bereichen mit relativ geringer Dichte der Datenverteilung liegen, in denen das Modell nicht ausreichend trainiert wurde. In dieser Arbeit, analysieren wir die Auswirkung verschiedener Arten von Rauschen auf die Vorhersagen von Anwendungen, die auf tiefen Neuronalen Netzen basieren. Insbesondere ein verallgemeinertes Rahmenwerk für die Erstellung negativer Beispiele in einer Blackbox-Angriffssituation für klassifikationsbasierte Modelle Präsentieren wir. Zum Schutz vor solchen schädlichen Beispielen entwickeln wir einen neuartigen Algorithmus namens MALADE, der den gegebenen Off-Manifold-Input in Richtung der Regionen mit hoher Dichte der datenerzeugenden Verteilung. Hierbei machen wir uns intrinsisches Wissen über die wahrnehmungsbezogene Entscheidungsgrenze während der Inferenz zunutze. Für auf Domänenübersetzung basierende Modelle, schlagen wir vor, Ergebnisse von niedriger Qualität in Richtung der Datenverteiler zu lenken durch Abkühlung der Eingabe-Test-Verteilung mit Hilfe der Langevin-Dynamik. Wir zeigen qualitativ und quantitativ, dass unsere Strategie die Robustheit von State-of-the-Art-Methoden für die Klassifikation sowie für Domänenübersetzungsaufgaben verbessert. Am Beispiel von medizinischen Bildgebung als exemplarischem Anwendungsfall Von Klassifikation basierend auf neuronalen Netzen bewerten wir die Robustheit von Pretraining und Selbstüberwachungsstrategien gegenüber Eingabeverzerrungen und systematischem Bias

    Viešojo sektoriaus darbuotojų kokybės nuostatos Lietuvoje ir Indijoje.

    No full text
    The quality of services provided by public sector employees is directly related to the quality of life perceived by the population. The quality of the services provided by public sector bodies and the underlying factors is an important research subject. This thesis examines differences between the quality of staff regulations in India and Lithuania and the quality of applied practices

    In-Situ Infrared Spectroscopic Studies of Palladium Thin Films during CO2 Electro-Reduction

    No full text
    An exponential growth in CO2 concentration over the past few decades has led to an accelerated impact of climate change on planet earth. In a bid to curb these emissions, people across the globe are slowly transitioning towards renewable energy sources with battery technology aiding this growth. Given that battery technology is still in its nascent stage, the “Electrochemical reduction of CO2” could be a viable solution supporting it without decelerating the momentum gained towards renewable development. Although plausible, the direct reduction of CO2 to liquid fuels entails huge energy expenditure thus requiring the implementation of catalysts. Unique in its ability, palladium reversibly reduces CO2 to formic acid making it an interesting candidate for the reduction reaction. In addition to the production of formic acid, palladium is also know to produce carbon monoxide (CO) which completely deactivates the surface preventing further reactions from occurring. Thus the aim of the current study is focused on analysing the electrochemical reduction of CO2 on palladium thin films using surface enhanced infrared absorption spectroscopy to better understand the deactivation mechanisms of CO on the palladium thin film. The smoothness of the as- sputtered 15 nm palladium thin film with a RMS roughness of 0.511 nm and partially coalesced islands were ascertained, thus requiring surface activation to introduce the enhancement mechanism. Experimental analysis of CO2 reduction on the palladium thin film was performed to unearth significant insights through the combination of electrochemical analysis techniques with surface enhanced infrared absorption spectroscopy. Results obtained through implementation of these methodologies provided substantial information not only on the influence of the palladium-hydrogen system on the electrochemical reduction of CO2 but also on the impact of alkali metal cations on the palladium-hydrogen system and the CO2 reduction reaction over the sputtered palladium thin film. CO formation, accumulation and desorption coupled with hydrogen evolution and desorption were some of the few avenues that were enumerated upon during the experimental investigation. The identity of CO chemisorbed on the palladium thin film along with bicarbonate direct/ indirect reduction to form CO was confirmed through the utilization of N2 saturated C13 NaHCO3 solution. In addition to the analysis of the reduction reaction, emphasis on the oxidation of CO was also provided suggesting the formation of dense CO structures with the existence of strong CO dipole – dipole coupling on the palladium surface. Materials Science and Engineerin

    Adversarial detection using denoising autoencoder

    No full text
    LAUREA MAGISTRALEGli Adversarial examples sono input creati maliziosamente per le reti neurali. Questi attacchi sono generati aggiungendo modifiche umanamente impercettibili alle immagini con l'obiettivo di ingannare il modello, causando quindi errori di classificazione. Per applicazioni mediche, auto a guida autonoma e software di riconoscimento facciale, questi esempi contraddittori rappresentano potenziali minacce. I metodi di Adversarial detection cercano di riconoscere queste perturbazioni. In questa tesi viene proposto uno di questi metodi basato sui Denoising Autoencoders (DAE). Il gradiente dell'energia, noto anche come score function dell'immagine di input, può essere stimato dalla ricostruzione del DAE. La score function di un adversial input può essere considerevolmente più grande della score function di un'immagine utilizzata durante il training. Quindi, con una soglia scelta in modo appropriato, gli Adversarial examples possono essere rilevati prima di essere passati al classificatore. Questa tesi esplora le condizioni per rendere il DAE robusto e mostra inoltre che il DAE supera il metodo stato dell'arte (SOTA) per fare detection.Adversarial examples are maliciously crafted inputs to neural networks. These attacks are generated by adding humanly imperceptible changes to images with the goal to fool the model, i.e., cause misclassification. For medical applications, self-driving cars, and facial recognition software these adversarial examples represent potential threats. Adversarial detection methods try to recognize these perturbations. In this thesis one such method is proposed based on Denoising Autoencoders (DAE). The gradient of the energy, also known as the score function of the input image can be estimated from reconstruction of the DAE. The score function of an adversarial input can be relatively larger to the score function of an image used during training. Hence, with an appropriately chosen threshold, the adversarial examples can be detected before being passed to the classifier. This thesis explores conditions favourable for making the DAE robust and also shows that DAE outperforms state-of-the-art (SOTA) baseline method for detection
    corecore