Search

Your search keyword '"Dumke, Markus"' showing total 9 results
Start Over Author "Dumke, Markus" Publication Year Range Last 50 years
9 results on '"Dumke, Markus"'

1. Double Q($\sigma$) and Q($\sigma, \lambda$): Unifying Reinforcement Learning Control Algorithms

Author

Dumke, Markus

Subjects
Computer Science - Artificial Intelligence, Computer Science - Learning, Statistics - Machine Learning
Abstract

Temporal-difference (TD) learning is an important field in reinforcement learning. Sarsa and Q-Learning are among the most used TD algorithms. The Q($\sigma$) algorithm (Sutton and Barto (2017)) unifies both. This paper extends the Q($\sigma$) algorithm to an online multi-step algorithm Q($\sigma, \lambda$) using eligibility traces and introduces Double Q($\sigma$) as the extension of Q($\sigma$) to double learning. Experiments suggest that the new Q($\sigma, \lambda$) algorithm can outperform the classical TD control methods Sarsa($\lambda$), Q($\lambda$) and Q($\sigma$).

Published
2017

3. Survival-Adjusted FEV1 and BMI Percentiles for Patients with Cystic Fibrosis before the Era of Triple CFTR Modulator Therapy in Germany.

Author

Dittrich, A. Susanne, Dumke, Markus, Kapl, Ferdinand, Schneider, Philipp, Wege, Sabine, Gräber, Simon, Stahl, Mirjam, Herth, Felix J., Naehrlich, Lutz, Mall, Marcus A., and Sommburg, Olaf

Subjects
*REPORTING of diseases, *REGRESSION analysis, *CYSTIC fibrosis, *FORCED expiratory volume, *RESEARCH funding, *BODY mass index, *MEMBRANE proteins, *LONGITUDINAL method, *EVALUATION
Abstract

Background: Pulmonary disease is the major cause for morbidity and mortality in cystic fibrosis (CF). In CF, forced expiratory volume in 1 s (FEV1) referenced against a healthy population (FEV1%predicted) and body mass index (BMI) do not allow for the comparison of disease severity across age and gender. Objectives: We aimed to determine updated FEV1 and BMI percentiles for patients with CF and to study their dependence on mortality attrition. Methods: Age- and height-adjusted FEV1 and BMI percentiles for CF patients aged 6–50 years were calculated from 4,947 patients of the German CF Registry for the period 2016–2019 utilizing quantile regression and a Generalized Additive Model for Location, Scale and Shape (GAMLSS). Further, survival-adjusted percentiles were estimated. Results: In patients with CF, FEV1 increased throughout childhood until maximal median values at 16 years in females (2.46 L) and 18 years in males (3.27 L). During adulthood, FEV1 decreased substantially. At 17 years of age, the 25th BMI percentile of patients with CF (females 18.50 and males 18.15 kg/m2) was below the 10th BMI percentile of the German reference cohort. From the age of 20 years, survival (96.3%) decreased tremendously. At 50 years of age (survival 15.0%), the 50th CF-specific FEV1 or BMI percentile among the survivors corresponded to the 92.5th percentile among the total CF birth cohort. Conclusions: Continuously updated disease-specific FEV1 and BMI percentiles with correction for survival may serve as age-independent measure of disease severity in CF (accessible via https://cfpercentiles.statup.solutions). [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

7. Double Q($��$) and Q($��, ��$): Unifying Reinforcement Learning Control Algorithms

Author

Dumke, Markus

Subjects
FOS: Computer and information sciences, Artificial Intelligence (cs.AI), Machine Learning (stat.ML), Machine Learning (cs.LG)
Abstract

Temporal-difference (TD) learning is an important field in reinforcement learning. Sarsa and Q-Learning are among the most used TD algorithms. The Q($��$) algorithm (Sutton and Barto (2017)) unifies both. This paper extends the Q($��$) algorithm to an online multi-step algorithm Q($��, ��$) using eligibility traces and introduces Double Q($��$) as the extension of Q($��$) to double learning. Experiments suggest that the new Q($��, ��$) algorithm can outperform the classical TD control methods Sarsa($��$), Q($��$) and Q($��$).

Published
2017
Full Text
View/download PDF

8. Reference Values Matter

Author

Ellemunter, Helmut, Dumke, Markus, and Steinkamp, Gratiana

Abstract

Reference values are important for patient care as well as for comparisons between different centers or countries. We investigated two anthropometric reference datasets, the US Centers for Disease Control (CDC) growth charts and the German Health Interview and Examination Survey for Children and Adolescents Study (KiGGS) percentiles, which were established in Germany between 2003 and 2006. A smaller proportion of children with cystic fibrosis had decreased zscores <-2 with CDC (5.0% for weight and 3.0% for height) compared to KiGGS (7.4% for weight and 6.3% for height) values (P< 0.0001). Median zscores were higher using the CDC reference data. Thus, the choice of growth reference is important, may influence clinical management and must be considered when comparing the outcomes of different institutions.

Published
2021
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results