|
|
Deep Learning Model Uncertainty in Medical Image Analysis
Drevický, Dušan ; Kolář, Martin (referee) ; Kodym, Oldřich (advisor)
Táto práca sa zaoberá určením neistoty v predikciách modelov hlbokého učenia. Aj keď sa týmto modelom darí dosahovať vynikajúce výsledky v mnohých oblastiach počítačového videnia, ich výstupy sú väčšinou deterministické a neposkytujú mnoho informácií o tom, ako si je model istý svojou predpoveďou. To je obzvlášť dôležité pri analýze lekárskych obrazových dát, kde môžu mať omyly vysokú cenu a schopnosť detekovať neisté predikcie by umožnila dohliadajúcemu lekárovi spracovať relevantné prípady manuálne. V tejto práci aplikujem niekoľko rôznych metrík vyvinutých v nedávnom výskume pre určenie neistoty na modely hlbokého učenia natrénované pre lokalizáciu cefalometrických landmarkov. Následne ich vyhodnotím a porovnávam v sade experimentov, ktorých úlohou je určiť, nakoľko jednotlivé metriky poskytujú užitočnú informáciu o tom, ako si je model istý svojou predpoveďou.
|
|
|
Generic Approach to Updating Uncertainty: Focus on Conditioning
Kuncová, Alexandra ; Peliš, Michal (advisor) ; Sedlár, Igor (referee)
First, we consider different kinds of representation of uncertainty and the meth- ods for updating each of them by conditioning. We focus on the generic frame- work of (conditional) plausibility spaces, since it generalises all the introduced representations. Further, we select three frameworks and list the properties that need to be added to a conditional plausibility space in order to recover each of these frameworks. The main goal of this work, however, is to show how public announcement on single-agent plausibility models, ranking structures, and pos- sibility structures realised by their corresponding update mechanisms, can be embedded into the framework of conditional plausibility spaces. At the very end we briefly illustrate a general update model using plausibility measures. Keywords: belief revision, dynamic logic, epistemic logic, plausibility space, pub- lic announcement, uncertainty, update.
|
|
|
The Inflation-Output Variability Relationship in the CEE countries: A Bivariate GARCH Model
Kubovič, Jozef ; Čech, František (advisor) ; Červinka, Michal (referee)
This thesis examines the output-variability relationship and causal relationships among the inflation, the output growth and their uncertainties for the Central and Eastern European region during the period of time that covers the economic crisis of 2008. We apply the bivariate GARCH(1,1) model with the constant conditional correlation covariance matrix to obtain conditional variances that proxy the two uncertainties and use Granger causality test to determine the causal effects among four variables. We come up with a number of interesting results. First, we did not find statistical evidence neither for the inflation-output variability relationship nor for the Phillips curve. Second, we uncovered support for the positive causal effect of the inflation on its uncertainty and negative causal effect for the reverse direction. Additionally, we also found some support for the indirect negative causal effect of the inflation on the output growth. These results support the policy of low and stable inflation in the countries. Finally, we showed that crisis has a significant impact on the results, changing the behaviour of conditional variances and causal effects among the variables. Powered by TCPDF (www.tcpdf.org)
|
|
|
Deep Learning Model Uncertainty in Medical Image Analysis
Drevický, Dušan ; Kolář, Martin (referee) ; Kodym, Oldřich (advisor)
Táto práca sa zaoberá určením neistoty v predikciách modelov hlbokého učenia. Aj keď sa týmto modelom darí dosahovať vynikajúce výsledky v mnohých oblastiach počítačového videnia, ich výstupy sú väčšinou deterministické a neposkytujú mnoho informácií o tom, ako si je model istý svojou predpoveďou. To je obzvlášť dôležité pri analýze lekárskych obrazových dát, kde môžu mať omyly vysokú cenu a schopnosť detekovať neisté predikcie by umožnila dohliadajúcemu lekárovi spracovať relevantné prípady manuálne. V tejto práci aplikujem niekoľko rôznych metrík vyvinutých v nedávnom výskume pre určenie neistoty na modely hlbokého učenia natrénované pre lokalizáciu cefalometrických landmarkov. Následne ich vyhodnotím a porovnávam v sade experimentov, ktorých úlohou je určiť, nakoľko jednotlivé metriky poskytujú užitočnú informáciu o tom, ako si je model istý svojou predpoveďou.
|
|
|
The Inflation-Output Variability Relationship in the CEE countries: A Bivariate GARCH Model
Kubovič, Jozef ; Čech, František (advisor) ; Červinka, Michal (referee)
This thesis examines the output-variability relationship and causal relationships among the inflation, the output growth and their uncertainties for the Central and Eastern European region during the period of time that covers the economic crisis of 2008. We apply the bivariate GARCH(1,1) model with the constant conditional correlation covariance matrix to obtain conditional variances that proxy the two uncertainties and use Granger causality test to determine the causal effects among four variables. We come up with a number of interesting results. First, we did not find statistical evidence neither for the inflation-output variability relationship nor for the Phillips curve. Second, we uncovered support for the positive causal effect of the inflation on its uncertainty and negative causal effect for the reverse direction. Additionally, we also found some support for the indirect negative causal effect of the inflation on the output growth. These results support the policy of low and stable inflation in the countries. Finally, we showed that crisis has a significant impact on the results, changing the behaviour of conditional variances and causal effects among the variables. Powered by TCPDF (www.tcpdf.org)
|
|
|
Generic Approach to Updating Uncertainty: Focus on Conditioning
Kuncová, Alexandra ; Peliš, Michal (advisor) ; Sedlár, Igor (referee)
First, we consider different kinds of representation of uncertainty and the meth- ods for updating each of them by conditioning. We focus on the generic frame- work of (conditional) plausibility spaces, since it generalises all the introduced representations. Further, we select three frameworks and list the properties that need to be added to a conditional plausibility space in order to recover each of these frameworks. The main goal of this work, however, is to show how public announcement on single-agent plausibility models, ranking structures, and pos- sibility structures realised by their corresponding update mechanisms, can be embedded into the framework of conditional plausibility spaces. At the very end we briefly illustrate a general update model using plausibility measures. Keywords: belief revision, dynamic logic, epistemic logic, plausibility space, pub- lic announcement, uncertainty, update.
|
guest ::
