In the thesis we analyse changes in derivatives valuation after the financial crisis and their impact on behaviour of financial institutions. We focus mainly on the changes related to counterparty credit risk and valuation adjustments. We describe in economical terms the relationship between counterparty credit risk and traditional credit risk, we also introduce management and modelling of this risk. In second part of the study we analyse the regulatory framework, in particular new capital requirement and mandatory central clearing of OTC derivatives. We discuss inconsistencies between regulatory and internal approaches to the counterparty risk measurement and also significant systemic risk connected to central counterparties. Finally we investigate the impact of changes in derivatives valuation on banks in both the EU and the Czech Republic. Specifically we are interested in optimal approach to entering into derivative trade.

Since the inverse problem in Diffusive Optical Tomography (DOT) is nonlinear and severely ill-posed, only low resolution reconstructions are feasible when noise is added to the data nowadays. The purpose of this thesis is to improve image reconstruction in DOT of the main optical properties of tissues with some novel mathematical methods. We have used the Landweber (L) method, the Landweber-Kaczmarz (LK) method and its improved Loping-Landweber-Kaczmarz (L-LK) method combined with sparsity or with total variation regularizations for single and simultaneous image reconstructions of the absorption and scattering coefficients. The sparsity method assumes the existence of a sparse solution which has a simple description and is superposed onto a known background. The sparsity method is solved using a smooth gradient and a soft thresholding operator. Moreover, we have proposed an improved sparsity method. For the total variation reconstruction imaging, we have used the split Bregman method and the lagged diffusivity method. For the total variation method, we also have implemented a memory-efficient method to minimise the storage of large Hessian matrices. In addition, an individual and simultaneous contrast value reconstructions are presented using the level set (LS) method. Besides, the shape derivative of DOT based on the RTE is derived using shape sensitivity analysis, and some reconstructions for the absorption coefficient are presented using this shape derivative via the LS method.\\Whereas most of the approaches for solving the nonlinear problem of DOT make use of the diffusion approximation (DA) to the radiative transfer equation (RTE) to model the propagation of the light in tissue, the accuracy of the DA is not satisfactory in situations where the medium is not scattering dominant, in particular close to the light sources and to the boundary, as well as inside low-scattering or non-scattering regions. Therefore, we have solved the inverse problem in DOT by the more accurate time-dependant RTE in two dimensions.

Cette thèse est composée de deux parties: une première partie traite le problème de changement de régime et une deuxième partie concerne le processusautorégressif à seuil dont les innovations ne sont pas indépendantes. Toutefois, ces deux domaines de la statistique et des probabilités se rejoignent dans la littérature et donc dans mon projet de recherche. Dans la première partie, nous étudions le problème de changements derégime. Il existe plusieurs méthodes pour la détection de ruptures mais les principales méthodes sont : la méthode de moindres carrés pénalisés (PLS)et la méthode de derivée filtrée (FD) introduit par Basseville et Nikirov. D’autres méthodes existent telles que la méthode Bayésienne de changementde points. Nous avons validé la nouvelle méthode de dérivée filtrée et taux de fausses découvertes (FDqV) sur des données réelles (des données du vent sur des éoliennes et des données du battement du coeur). Bien naturellement, nous avons donné une extension de la méthode FDqV sur le cas des variables aléatoires faiblement dépendantes.Dans la deuxième partie, nous étudions le modèle autorégressif à seuil (en anglais Threshold Autoregessive Model (TAR)). Le TAR est étudié dans la littérature par plusieurs auteurs tels que Tong(1983), Petrucelli(1984, 1986), Chan(1993). Les applications du modèle TAR sont nombreuses par exemple en économie, en biologie, l'environnement, etc. Jusqu'à présent, le modèle TAR étudié concerne le cas où les innovations sont indépendantes. Dans ce projet, nous avons étudié le cas où les innovations sont non corrélées. Nous avons établi les comportements asymptotiques des estimateurs du modèle. Ces résultats concernent la convergence presque sûre, la convergence en loi et la convergence uniforme des paramètres. / This thesis has two parts: the first part deals the change points problem and the second concerns the weak threshold autoregressive model (TAR); the errors are not correlated.In the first part, we treat the change point analysis. In the litterature, it exists two popular methods: The Penalized Least Square (PLS) and the Filtered Derivative introduced by Basseville end Nikirov.We give a new method of filtered derivative and false discovery rate (FDqV) on real data (the wind turbines and heartbeats series). Also, we studied an extension of FDqV method on weakly dependent random variables.In the second part, we spotlight the weak threshold autoregressive (TAR) model. The TAR model is studied by many authors such that Tong(1983), Petrucelli(1984, 1986). there exist many applications, for example in economics, biological and many others. The weak TAR model treated is the case where the innovations are not correlated.

Este trabalho foi concebido inicialmente através de projeto, montagem e ensaios preliminares de uma unidade analógica, para propósito didático, composta de um servomecanismo controlado pelas clássicas estratégias Proporcional, Integral e Derivativa (PID). Com este equipamento foi demonstrado os conceitos fundamentais da engenharia de controle com seus ensaios e medidas que permitem exemplificar, de forma simples e bastante pedagógica, o funcionamento e os limites operacionais de um controlador prático. Os valores medidos em experimento foram comparados com valores obtidos pela simulação numérica de um modelo equivalente implementado no software MatLab. Utilizando-se da tecnologia PIC (Controlador de Interface Programável), o hardware analógico foi substituído, com a intenção de comparar os resultados para melhor ilustrar os abstratos conceitos da ação PID sobre um dispositivo eletromecânico razoavelmente linear. A unidade desenvolvida permite estabelecer um elo valioso à aprendizagem, buscando unir a teoria clássica, a simulação numérica, a aplicação digital e o comportamento real em uma situação física. Tivemos êxito nos resultados, conseguindo comprovar as modelagens matemáticas em aplicações práticas, além de poder comparar os desempenhos dos resultados analógicos com os resultados digitais. / This work was initially conceived by design, assembly and preliminary tests of an analogic unit, for didactic purpose, composed of a servomechanism controlled by the classic strategies Proportional, Integral and Derivative (PID). With this device was demonstrated the fundamental concepts of control engineering and its trials and measures allow exemplify, in a simple and very interactive way, the functioning and the operational limits of a practical controller. The values measured in assay were compared with values obtained by numerical simulation of an equivalent model implemented in MatLab. By using the technology PIC (Programmable Interface Controller), the analogical hardware will be substituted, in order to compare the results to better explain the abstract concepts of the PID action over a sufficiently linear electromechanical device. The developed unit enables a valuable link to the learning, seeking to unite the classical theory, numerical simulation, digital application and real performance in a physical case. We were successful in results, obtaining proof of the mathematical models in practical applications, besides being able to compare the performances of analogic results with digital ones.

O presente trabalho teve como objetivo avaliar níveis de 0, 3, 6 e 9% de glicerol na dieta de suínos em crescimento e terminação por meio do desempenho, das características de carcaça e da qualidade da carne. Foram utilizados 64 animais da genética Topigs com peso médio inicial 33,27 ± 4,66 kg, distribuídos em 32 baias de acordo com o sexo e peso inicial, em um delineamento em blocos casualizados com oito repetições (blocos) por tratamento. Em cada uma das três fases, crescimento I (33,27 a 65,00 kg), crescimento II (65,00 a 85,00 kg) e terminação (85,00 a 99,97 kg), os animais receberam rações isonutritivas e água à vontade. Ao atingirem o peso vivo médio de 99,97 ± 1,92 kg, os animais foram abatidos e as carcaças avaliadas para o rendimento de carcaça quente, comprimento de carcaça, espessura de toicinho, área de olho de lombo e calculada a relação gordura/carne. Amostras do músculo Longissimus dorsi foram retiradas para medição do pH, cor e perda de água por gotejamento. A adição de glicerol proporcionou redução apenas no ganho diário de peso (P=0,04) durante as fases de crescimento I e II. De modo geral, o glicerol, um subproduto da produção de biodiesel, pode ser utilizado como ingrediente energético de rações de suínos em crescimento e terminação até o nível de 9%, sem afetar sensivelmente o desempenho, as características de carcaça e a qualidade da carne dos animais. / The purpose of this work was to evaluate the effect of 0, 3, 6 and 9% of glycerol in growing and finishing pig diets on performance, carcass traits and meat quality. Sixty-four pigs of Topigs genetic with 33.27 ± 4.66 kg initial live weight were allotted to 32 pens, according to sex and initial weight in a randomized complete block design with eight replications (blocks) per treatment. In each of three phases, growing I (33.27 to 65.00 kg), growing II (65.00 to 85.00 kg) and finishing (85.00 to 99.97 kg), diets and water were given ad libitum to animals. When pigs reached 99.97 ± 1.92 kg live weight, they were slaughtered and data of hot carcass yield, carcass length, backfat thickness, loin eye area and fat area/loin eye area ratio were registered. Samples of Longissimus dorsi were taken to determine pH, colour and water drip loss. Added glycerol depressed average daily gain (P=0.04) only during growing I and II periods. Overall, glycerol, a by-product of biodiesel production, can be used as energy source of growing and finishing pig diets up to 9% without affecting performance, carcass traits and meat quality of animals.

Cette thèse est consacrée à l’étude du calcul fonctionnel non-anticipatif, qui est basé sur la notion de dérivée verticale d'une fonctionelle. Nous étendons le cadre classique de ce calcul à des fonctionnelles ne possédant pas de dérivée directionnelle classique. Dans la première partie, nous montrons comment une classe importante de fonctionelles, définie par une espérance conditionnelle, peuvent être approchées de façon systématique par des fonctionnelles régulières. Dans la deuxième partie, nous introduisons une notion de dérivée verticale faible qui couvre une plus grande classe de fonctionnelles, et notamment toutes les martingales locales. Dans la première partie, nous nous sommes intéressés à la représentation d'une espérance conditionnelle par une fonctionnelle non-anticipative. L'idée est d'approximer ces fonctionnelles par une suite des fonctionnelles régulières dans un certain sens. Cette approche fournit une façon systématique d'obtenir une approximation explicite de la représentation des martingales pour une grande famille de fonctionnelles Browniennes. Nous obtenons également un ordre de convergence explicite. Quelques applications au problème de la couverture dynamique sont données à la fin de cette partie.Dans la deuxième partie, nous étendons la notion de dérivée verticale pour des fonctionnelles qui n'admettent pas nécessairement de dérivée directionnelle. Cette notion nous permet également d'obtenir une caractérisation fonctionnelle d'une martingale locale par rapport à un processus de référence fixé, ce qui donne lieu à une notion de solution faible pour des équations aux dérivées partielles dépendant de la trajectoire. / This thesis focuses on various mathematical questions arising in the non-anticipative functional calculus, which is based on a notion of pathwise directional derivatives for functionals. We extend the scope and results of this calculus to functionals which may not admit such derivatives, either through approximations (Part I) or by defining a notion of weak vertical derivative (Part II). In the first part, we consider the representation of conditional expectations as non-anticipative functionals. We show that it is possible under very general conditions to approximate such functionals by a sequence of smooth functionals in an appropriate sense. This approach provides a systematic method for computing explicit approximations to martingale representations for a large class of Brownian functionals. We also derive explicit convergence rates of the approximations. These results are then applied to the problem of sensitivity analysis and dynamic hedging of (path-dependent) contingent claims. In the second part, we propose a concept of weak vertical derivative for non-anticipative functionals which may fail to possess directional derivatives. The definition of the weak vertical derivative is based on the notion of pathwise quadratic variation and makes use of the duality associated to the associated bilinear form. We show that the notion of weak vertical derivative leads to a functional characterization of local martingales with respect to a reference process, and allows to define a concept of pathwise weak solution for path-dependent partial differential equations.

Les anticorps monoclonaux (Acms) représentent une part importante des traitements médicamenteux à visées anticancéreuses à l'hôpital. Leur action ciblée et leur efficacité favorisent leur essor considérable dans le monde pharmaceutique. Leur reconstitution est nécessaire avant d'être administré par voie intraveineuse chez le patient. A l’heure actuelle, il n’existe pas de contrôle rigoureux permettant leur indentification. Compte-tenu de leurs propriétés physico-chimiques similaires leur discrimination par un contrôle analytique hospitalier pré-libératoire est un véritable défi. Dans cette thèse trois nouvelles approches ont été explorées pour permettre la discrimination et l'identification rapide d’un mélange de quatre anticorps monoclonaux après reconstitution: le bevacizumab, le cetuximab, le rituximab et le trastuzumab. Parmi ces approches, deux portent sur des méthodes séparatives : la chromatographie d'échange de cations sur support monolithique et l'électrophorèse capillaire de zone. Les conditions d'analyse de ces méthodes ont été optimisées dans l'objectif d’une identification rapide et fiable. Nous avons également étudié les paramètres intrinsèques des anticorps monoclonaux influençant leur séparation. La troisième approche est basée sur une méthode spectrale ultra-rapide par spectroscopie UV dérivée associant l'analyse par injection en flux et l'analyse multivariée. L'étude des données spectrales par l'analyse multivariée nous a permis de concevoir un algorithme capable de discriminer sans ambigüité les anticorps monoclonaux à partir de leurs informations spectrales. / Anticancer monoclonal antibodies (mAbs) represent a large part of anticancer drugs at hospital. Their success is related to their capacity to specifically interact with their target with relatively low adverse effects. mAbs must be compounded at the hospital before patient's infusion. They exhibit common features rendering their discrimination through fast analytical methods difficult. At the hospital, quality controls mainly rely on mAbs identity and quantity compliance. However, identity of compounds with high similarity remains challenging. In this thesis, three new approaches for the discrimination and the identification of four mAbs, bevacizumab, cetuximab, rituximab and trastuzumab have been investigated. Among them, two approaches are based on separation methods Cation exchange chromatography on a monolithic stationary phase and capillary electrophoresis. The analytical conditions were extensively optimized to obtain rapid and suitable method allowing their discrimination. We have also studied the influence of their physico-chemical properties on their separations. The third approach relies on a rapid second derivative spectral method combining flow injection analysis and multivariate analysis. An algorithm able to discriminate accurately the monoclonal antibodies has been designed from in-depth study of the spectral data through multivariate analysis.

<p>U&nbsp; doktorskoj&nbsp; disertaciji&nbsp; je&nbsp; prezentovano&nbsp; pedago&scaron;ko&nbsp; istraživanje&nbsp; koje&nbsp; se&nbsp; odnosi&nbsp; na&nbsp; teorijsko&nbsp; i eksperimentalno&nbsp; ispitivanje&nbsp; efekata&nbsp; primene&nbsp; metodskih&nbsp; pristupa&nbsp; zasnovanih&nbsp; na&nbsp; matematičkom modelovanju u obradi izvoda funkcije i njegove primene u visokom strukovnom obrazovanju. Na osnovu teorijskih principa na kojima je zasnovan proces matematičkog modelovanja, osmi&scaron;ljen je način za implementaciju modelovanja u nastavni proces i kreirani su modeli za realizaciju obrade sadržaja iz oblasti izvoda funkcije i njegove primene. Predložen je i originalni pristup matematičkom modelovanju koji se realizuje u računarskom okruženju i istaknute su sve prednosti novog pristupa<br />koje&nbsp; se&nbsp; tiču&nbsp; realizacije&nbsp; nastavnog&nbsp; procesa&nbsp; i&nbsp; rezultata&nbsp; učenja&nbsp; i&nbsp; poučavanja.&nbsp; Disertacija&nbsp; se&nbsp; bavi&nbsp; i savremenim trendovima u obrazovanju nastavnika i njegovom unapređenju. Poseban akcenat je stavljen&nbsp; na&nbsp; osmi&scaron;ljavanje&nbsp; elemenata&nbsp; obuke&nbsp; nastavnika&nbsp; za&nbsp; primenu&nbsp; matematičkog&nbsp; modelovanja&nbsp; u &scaron;kolskoj praksi.<br />Istraživanje o efektima primene matematičkog modelovanja je sprovedeno u dva ciklusa. Prvi ciklus istraživanja je realizovan u periodu od 2009-2014. godine sa četiri generacije studenata strukovnih studija. U eksperimentu je učestvovalo ukupno 555 studenata organizovanih u paralelne grupe. U prvom&nbsp; ciklusu&nbsp; istraživanja&nbsp; je&nbsp; praćen&nbsp; uticaj&nbsp; primene&nbsp; matematičkog&nbsp; modelovanja&nbsp; na&nbsp; postignuća studenata iz oblasti izvoda funkcije i njegove primene. Od instrumenata primenjenih za ispitivanje postignuća&nbsp; studenata&nbsp; su&nbsp; kori&scaron;ćeni&nbsp; testovi&nbsp; znanja&nbsp; (kolokvijum&nbsp; i&nbsp; ispit)&nbsp; i&nbsp; anketa&nbsp; koja&nbsp; je&nbsp; ispitivala stavove studenata o realizaciji nastave matematike i njenoj korisnosti.<br />Drugi&nbsp; ciklus&nbsp; istraživanja&nbsp; je&nbsp; sproveden&nbsp; &scaron;kolske&nbsp; 2016/2017.&nbsp; godine.&nbsp; Eksperiment&nbsp; je&nbsp; realizovan&nbsp; sa paralelnim grupama u kojima su učestvovala 204 studenta Visoke&nbsp; tehničke &scaron;kole&nbsp; strukovnih&nbsp; studija.U&nbsp; drugom&nbsp; ciklusu&nbsp; istraživanja&nbsp; su&nbsp; ispitani&nbsp; efekti&nbsp; primene&nbsp; novog&nbsp; pristupa&nbsp; matematičkom modelovanju&nbsp; u&nbsp; računarskom&nbsp; okruženju&nbsp; na&nbsp; znanja&nbsp; studenata&nbsp; iz&nbsp; oblasti&nbsp; izvoda&nbsp; funkcije&nbsp; i&nbsp; njegovih primena (eksperimentalna grupa) i dobijeni rezultati su&nbsp; upoređeni sa rezultatima koje su studenti ostvarili&nbsp; kada&nbsp; je&nbsp; nastava&nbsp; realizovana&nbsp; primenom&nbsp; tradicionalnog&nbsp; ciklusa&nbsp; modelovanja&nbsp; (kontrolna<br />grupa).Na&nbsp; osnovu&nbsp; rezultata&nbsp; oba&nbsp; pedago&scaron;ka&nbsp; istraživanja,&nbsp; utvrđeno&nbsp; je&nbsp; da&nbsp; realizacija&nbsp; nastave&nbsp; matematike primenom matematičkog modelovanja, kao i matematičkog modelovanja u računarskom okruženju, na obradu pojma izvoda funkcije i njegovih primena&nbsp; ima značajan uticaj na kvalitet znanja studenata i ostvarenost optimalnih rezultata u učenju, razumevanju nastavnih sadržaja i nji hovoj primeni na re&scaron;avanje problema iz ove oblasti.</p> / <p>In the doctoral dissertation, pedagogical research is presented, which refers to the theoretical and experimental examination of the effects of the application of methodological approaches based on mathematical modeling in the field of the&nbsp; derivative of the function and its application in college education.<br />Based on the theoretical principles on which the process of mathematical modeling is based, a way has been devised for the implementation of modeling in the teaching process and models have been created for the teaching process in the field of function &nbsp; derivative and its applications. An original approach to mathematical modeling that is realized in a computer environment is also proposed, and all the advantages of the new &nbsp; approach concerning the realization of the teaching process and the results of learning and teaching are highlighted. The dissertation also deals with modern trends in teacher education and its improvement. Special emphasis is placed on designing elements&nbsp; of&nbsp; teacher&nbsp; training&nbsp; for&nbsp; the&nbsp; application&nbsp; of&nbsp; mathematical&nbsp; modeling&nbsp; in&nbsp; school&nbsp; practice.<br />The research on the effects of the application of mathematical modeling was conducted &nbsp; in two cycles. The first cycle of research was realized in the period from 2009-2014 with&nbsp; four generations of students. A total of 555 students organized in parallel groups participated in the experiment. In the&nbsp; first&nbsp; cycle&nbsp; of&nbsp; research,&nbsp; the&nbsp; influence&nbsp; of&nbsp; the&nbsp; application&nbsp; of&nbsp; mathematical&nbsp; modeling&nbsp; on&nbsp; the achievements of students in the field of function derivative and its applications was examined. Among the instruments used to examine students achievements, knowledge tests (colloquium and final&nbsp; exam)&nbsp; and&nbsp; a&nbsp; survey&nbsp; that&nbsp; examined&nbsp; students&#39;&nbsp; attitudes&nbsp; toward&nbsp; the&nbsp; teaching&nbsp; process&nbsp; and usefulness of mathematics were created and used.<br />The second cycle of research was conducted in the 2016/2017 school year. The experiment was realized&nbsp; with&nbsp; parallel&nbsp; groups&nbsp; in&nbsp; which&nbsp; participated&nbsp; 204&nbsp; students&nbsp; of&nbsp; the&nbsp; Technical&nbsp; College&nbsp; of Applied Sciences. In the second cycle of the research, the effects of applying a new approach to mathematical&nbsp; modeling&nbsp; in&nbsp; the&nbsp; computer&nbsp; environment&nbsp; on&nbsp; students&#39;&nbsp; knowledge&nbsp; of&nbsp; function derivative and its applications (experimental group) were examined and the obtained results were compared&nbsp; with&nbsp; the&nbsp; results&nbsp; achieved&nbsp; by&nbsp; students&nbsp; using&nbsp; traditional&nbsp; modeling&nbsp; cycle&nbsp; (the&nbsp; control group).<br />Based&nbsp; on&nbsp; the&nbsp; results&nbsp; obtained&nbsp; from&nbsp; both&nbsp; pedagogical&nbsp; researches,&nbsp; it&nbsp; was&nbsp; determined&nbsp; that&nbsp; the realization of teaching mathematics by applying mathematical modeling, as well as mathematical modeling&nbsp; in&nbsp; the&nbsp; computer&nbsp; environment,&nbsp; has&nbsp; a&nbsp; significant&nbsp; impact&nbsp; on&nbsp; the&nbsp; quality&nbsp; of&nbsp; students&#39; knowledge and the realization of optimal learning outcomes, and their application on solving the problems from this area.</p>

When a stock is added into the S&P 500 Index, it is automatically "cross-listed" in the index derivative markets (i.e., S&P 500 Index futures and Index options). I examined the effects of such cross-listing on the trading volume and return volatility of the underlying component stocks. Traditional finance theory asserts that futures and "cash" markets are connected by arbitrage mechanism that brings both markets to equilibrium. When arbitrage opportunities arise, arbitrageurs buy (sell) the index portfolio and take short (long) positions in the corresponding index derivative contracts until prices return to theoretical levels. Such mechanical arbitrage trading tends to create large order flows that could be difficult for the market to absorb, resulting in price changes. Utilizing a list of S&P 500 index composition changes occurring over the period September 1976 to December 2005, I investigated the market-adjusted volume turnover ratios and return variances of the stocks being added to and deleted from the S&P 500, surrounding the effective day of index membership changes. My primary finding is that, after the introduction of the S&P 500 index futures and options contracts, stocks added to the S&P 500 experience significant increase in both trading volume and return volatility. However, deleted stocks experience no significant change in either trading volume or return volatility. Both daily and monthly return variances increase following index inclusion, consistent with the hypothesis that derivative transactions "fundamentally" destabilize the underlying securities. I argue that the increase in trading volume and return volatility may be attributed to index arbitrage transactions as derivative markets provide more routes for index arbitrageurs to trade. Other index trading strategies such as portfolio insurance and program trading may also contribute to the results. On the other hand, a deleted stock is not associated with changes in trading volume and volatility since it represents an extremely small fraction of the market value-weighted index portfolio, and the influence of index trading strategies becomes slight for these shares. Furthermore, evidence is provided that trading volume and return volatility are positively related.

Monte Carlo methods in derivative pricing are computationally expensive, in particular for evaluating models partial derivatives with regard to inputs. This research proposes the use of deep learning to approximate such valuation models for highly exotic derivatives, using automatic differentiation to evaluate input sensitivities. Deep learning models are trained to approximate Phoenix Autocall valuation using a proprietary model used by Svenska Handelsbanken AB. Models are trained on large datasets of low-accuracy (10^4 simulations) Monte Carlo data, successfully learning the true model with an average error of 0.1% on validation data generated by 10^8 simulations. A specific model parametrisation is proposed for 2-day valuation only, to be recalibrated interday using transfer learning. Automatic differentiation approximates sensitivity to (normalised) underlying asset prices with a mean relative error generally below 1.6%. Overall error when predicting sensitivity to implied volatililty is found to lie within 10%-40%. Near identical results are found by finite difference as automatic differentiation in both cases. Automatic differentiation is not successful at capturing sensitivity to interday contract change in value, though errors of 8%-25% are achieved by finite difference. Model recalibration by transfer learning proves to converge over 15 times faster and with up to 14% lower relative error than training using random initialisation. The results show that deep learning models can efficiently learn Monte Carlo valuation, and that these can be quickly recalibrated by transfer learning. The deep learning model gradient computed by automatic differentiation proves a good approximation of the true model sensitivities. Future research proposals include studying optimised recalibration schedules, using training data generated by single Monte Carlo price paths, and studying additional parameters and contracts.