Clarke A.M.,Ecole Polytechnique Federale de Lausanne |
Friedrich J.,University of Bern |
Friedrich J.,Columbia University |
Tartaglia E.M.,French National Center for Scientific Research |
And 5 more authors.
PLoS ONE | Year: 2015
Humans can learn under a wide variety of feedback conditions. Reinforcement learning (RL), where a series of rewarded decisions must be made, is a particularly important type of learning. Computational and behavioral studies of RL have focused mainly on Markovian decision processes, where the next state depends on only the current state and action. Little is known about non-Markovian decision making, where the next state depends on more than the current state and action. Learning is non-Markovian, for example, when there is no unique mapping between actions and feedback. We have produced a model based on spiking neurons that can handle these non-Markovian conditions by performing policy gradient descent . Here, we examine the model's performance and compare it with human learning and a Bayes optimal reference, which provides an upper-bound on performance. We find that in all cases, our population of spiking neurons model well-describes human performance. © 2015 Clarke et al. Source
Beaussier H.,Groupe hospitalier Paris Saint Joseph |
Boutouyrie P.,Center dInvestigation Clinique |
Boutouyrie P.,UniversiteParis Descartes |
Boutouyrie P.,French Institute of Health and Medical Research |
And 8 more authors.
Journal of Hypertension | Year: 2015
Objectives: We assessed the influence of medication adherence on blood pressure (BP) control and target organ damage in a pre-specified analysis of a published trial comparing sequential nephron blockade (SNB) or sequential renin-angiotensin system blockade (SRASB) in patients with resistant hypertension. Methods: Patients were randomized to SNB (n=82) or SRASB (n=82) and studied at baseline and after 12 weeks. BP was measured by ambulatory blood pressure monitoring. Carotid-femoral pulse wave velocity (PWV) was measured by applanation tonometry and left ventricular mass (LVM) by echocardiography. Low medication adherence was assessed through plasma irbesartan concentration below 20 ng/ml; urinary N-acetylseryl-aspartyl-lysyl-proline/creatinine ratio below 4 nmol/ mmol; last medication intake before visit greater than 24 h and pill counting below 80% of theoretical intake. Medication adherence score (sum of items, max=4) is defined as low (medication adherence score <2) or acceptable (medication adherence score-2). Results: Among 164 patients, 134 (81.7%) had acceptable medication adherence and 30 (18.3%) low medication adherence, with similar proportions in the SNB and SRASB arms. After 12 weeks, in patients with acceptable medication adherence, BP was more frequently controlled in those treated with SNB (64%), than SRASB (18%; P<0.001). The difference in daytime SBP was-11.5mmHg [95% confidence interval (CI)-15.4 to-7.5, P<0.0001] in patients with acceptable medication adherence. In contrast, in patients with low medication adherence, the difference between groups was smaller and not significant (-9.4mmHg, 95% CI-20.4 to 1.7, P=0.09). Independently of BP changes, PWV and LVM decreased more in the SNB than in the SRASB arm when medication adherence was acceptable (-0.52 m/s, 95% CI-1.3 to-0.007, P=0.047; and-24 g/m2, 95% CI-36 to-12, P=0.0003), whereas no significant changes were observed in low medication adherence patients. Conclusion: Medication adherence contributes to BPlowering and regression of target organ damage. The differential effects of SNB and SRASB is observed in patients with acceptable medication adherence, and not in patients with low medication adherence © 2015 Wolters Kluwer Health, Inc. All rights reserved. Source