A Practical Approach to High-Performance Computing

Kurgalin, Sergei; Borzunov, Sergei

doi:10.1007/978-3-030-27558-7

Cited by 9 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Fourth, if, in addition to p-box parameters, some parameters are characterized by their precise CDFs, the optimization step is embedded in a Monte Carlo sampling loop ( S0 and S8 ); thereby increasing the number of optimizations by a factor of N (the total number of Monte Carlo samples). To decrease the computational cost, practitioners may opt to use more efficient optimization methods [41], fast-to-evaluate approximations of the original model or meta-models [42], parallelization of step S6 by distributing the optimization task across computing units (e.g., central processing unit cores) in combination with using high-performance computing [43], and a more efficient design-of-experiment for steps S3 and S4. [44] Nevertheless, we expect a higher computational burden since a PBA imposes fewer restrictions (i.e., we do not assume a functional form), leading to a larger region of uncertainty over which a model needs to be evaluated.…”

Section: Discussionmentioning

confidence: 99%

An approach to quantify parameter uncertainty in early assessment of novel health technologies

Iskandar

Federici

Berns

et al. 2022

Preprint

View full text Add to dashboard Cite

Health economic modeling of novel technology at the early stages of a product lifecycle has been used to identify technologies that are likely to be cost-effective. Such early assessments are challenging due to the potentially limited amount of data. Modelers typically conduct uncertainty analyses to evaluate their effect on decision-relevant outcomes. Current approaches, however, are limited in their scope of application and imposes an unverifiable assumption, i.e., uncertainty can be precisely represented by a probability distribution. In the absence of reliable data, an approach that uses the fewest number of assumptions is desirable. This study introduces a generalized approach for quantifying parameter uncertainty, i.e., probability bound analysis (PBA), that does not require a precise specification of a probability distribution in the context of early-stage health economic modeling. We introduce the concept of a probability box (p-box) as a measure of uncertainty without necessitating a precise probability distribution. We provide formulas for a p-box given data on summary statistics of a parameter. We describe an approach to propagate p-boxes into a model and provide step-by-step guidance on how to implement PBA. We conduct a case and examine the differences between the status-quo and PBA approaches and their potential implications on decision-making.

show abstract

Section: Discussionmentioning

confidence: 99%

An approach to quantify parameter uncertainty in early assessment of novel health technologies

Iskandar

Federici

Berns

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Fourth, if, in addition to p‐box parameters, some parameters are characterized by their precise CDFs, the optimization step is embedded in a Monte Carlo sampling loop ( S0 and S8 ); thereby increasing the number of optimizations by a factor of N (the total number of Monte Carlo samples). To decrease the computational cost, practitioners may opt to use more efficient optimization methods (Deng et al., 2013), fast‐to‐evaluate approximations of the original model or meta‐models (Ellis et al., 2020), parallelization of step S6 by distributing the optimization task across computing units (e.g., central processing unit cores) in combination with using high‐performance computing (Kurgalin & Borzunov, 2019), and a more efficient design‐of‐experiment for steps S3 and S4 (Schöbi & Sudret, 2017). Nevertheless, we expect a higher computational burden since a PBA imposes fewer restrictions (i.e., we do not assume a functional form), leading to a larger region of uncertainty over which a model needs to be evaluated.…”

Section: Discussionmentioning

confidence: 99%

An approach to quantify parameter uncertainty in early assessment of novel health technologies

et al. 2022

View full text Add to dashboard Cite

Health economic modeling of novel technology at the early stages of a product lifecycle has been used to identify technologies that are likely to be cost‐effective. Such early assessments are challenging due to the potentially limited amount of data. Modelers typically conduct uncertainty analyses to evaluate their effect on decision‐relevant outcomes. Current approaches, however, are limited in their scope of application and imposes an unverifiable assumption, that is, uncertainty can be precisely represented by a probability distribution. In the absence of reliable data, an approach that uses the fewest number of assumptions is desirable. This study introduces a generalized approach for quantifying parameter uncertainty, that is, probability bound analysis (PBA), that does not require a precise specification of a probability distribution in the context of early‐stage health economic modeling. We introduce the concept of a probability box (p‐box) as a measure of uncertainty without necessitating a precise probability distribution. We provide formulas for a p‐box given data on summary statistics of a parameter. We describe an approach to propagate p‐boxes into a model and provide step‐by‐step guidance on how to implement PBA. We conduct a case and examine the differences between the status‐quo and PBA approaches and their potential implications on decision‐making.

show abstract

“…Se espera construir y probar algoritmos de procesamiento aplicando computación paralela capaz de reducir los tiempos derivados de entrenar y ejecutar el modelo (Verner, Schuster & Silberstein, 2011;Verner, Schuster, Silberstein & Mendelson, 2012). Para ello es deseable facilitar el desarrollo de una solución paralela portable, de costo predecible, capaz de explotar las ventajas de modernos ambientes HPC a través de herramientas y "frameworks de computación" de alto nivel (Kurgalin & Borzunov, 2019;Pacheco, 2011).…”

Section: Desarrollounclassified

Analítica Prescriptiva en Sistemas con Grandes Flujos de Eventos

Schab

Piccoli²,

Casanova³

2020

AJEA

View full text Add to dashboard Cite

Los procesos de negocio exigen tomar decisiones rápidas para lograr la adaptación constante a los cambios en búsqueda de mejorar el desempeño y aprovechar las oportunidades. Una posibilidad es resolver el proceso computacionalmente. Para ello es necesario producir analíticas que transformen los datos en conocimiento para la toma de decisiones. Existen varios tipos de analíticas, en este trabajo se introduce una línea de investigación enfocada en la analítica prescriptiva, como nivel más avanzado, capaz de calcular acciones a ser ejecutadas en el momento (decisiones operativas) o en el futuro (decisiones tácticas para corto y mediano plazo, decisiones estratégicas para largo plazo) para lograr un objetivo deseado. El cálculo de las acciones involucra el procesamiento del flujo de eventos del negocio en forma de datastreams, la aplicación de técnicas y algoritmos de Soft Computing e Inteligencia Computacional y, derivado de la necesidad de bajos tiempos de respuesta, el empleo de Computación de Alto Desempeño.

show abstract

A Practical Approach to High-Performance Computing

Cited by 9 publications

References 0 publications

An approach to quantify parameter uncertainty in early assessment of novel health technologies

An approach to quantify parameter uncertainty in early assessment of novel health technologies

An approach to quantify parameter uncertainty in early assessment of novel health technologies

Analítica Prescriptiva en Sistemas con Grandes Flujos de Eventos

Contact Info

Product

Resources

About