1,393 research outputs found
Tight Performance Guarantees of Imitator Policies with Continuous Actions
No full textBehavioral Cloning (BC) aims at learning a policy that mimics the behavior demonstrated by an expert. The current theoretical understanding of BC is limited to the case of finite actions. In this paper, we study BC with the goal of providing theoretical guarantees on the performance of the imitator policy in the case of continuous actions. We start by deriving a novel bound on the performance gap based on Wasserstein distance, applicable for continuous-action experts, holding under the assumption that the value function is Lipschitz continuous. Since this latter condition is hardy fulfilled in practice, even for Lipschitz Markov Decision Processes and policies, we propose a relaxed setting, proving that value function is always H\"older continuous. This result is of independent interest and allows obtaining in BC a general bound for the performance of the imitator policy. Finally, we analyze noise injection, a common practice in which the expert’s action is executed in the environment after the application of a noise kernel. We show that this practice allows deriving stronger performance guarantees, at the price of a bias due to the noise addition
Delayed Reinforcement Learning by Imitation
No full textWhen the agent's observations or interactions are delayed, classic
reinforcement learning tools usually fail. In this paper, we propose a simple
yet new and efficient solution to this problem. We assume that, in the
undelayed environment, an efficient policy is known or can be easily learned,
but the task may suffer from delays in practice and we thus want to take them
into account. We present a novel algorithm, Delayed Imitation with Dataset
Aggregation (DIDA), which builds upon imitation learning methods to learn how
to act in a delayed environment from undelayed demonstrations. We provide a
theoretical analysis of the approach that will guide the practical design of
DIDA. These results are also of general interest in the delayed reinforcement
learning literature by providing bounds on the performance between delayed and
undelayed tasks, under smoothness conditions. We show empirically that DIDA
obtains high performances with a remarkable sample efficiency on a variety of
tasks, including robotic locomotion, classic control, and trading
Update of the budget impact analysis of the simplification to atazanavir + ritonavir + lamivudine dual therapy of HIV-positive patients receiving atazanavir-based triple therapies in Italy starting from data of the Atlas-M trial
No full textUmberto Restelli,1,2 Massimiliano Fabbiani,3 Simona Di Giambenedetto,3 Carmela Nappi,4 Davide Croce,1,21Center for Health Economics, Social and Health Care Management, LIUC – Università Cattaneo, Castellanza, Italy; 2School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa; 3Institute of Clinical Infectious Diseases, Catholic University of Sacred Heart, 4Health Economics, Bristol Myers Squibb S.r.l., Rome, ItalyIn 2017, the authors published an article to assess the financial consequences for the Italian National Health Service, over a 5-year period, of the adoption of a simplification strategy to atazanavir (ATV) + ritonavir (r) + lamivudine (3TC) dual therapy of HIVpositive patients receiving ATV plus two nucleoside reverse transcriptase inhibitors (NRTIs) starting from data of the Atlas-M trial at 48 weeks.1Consequently to the publication of the clinical results of the Atlas-M trial at 96 weeks, we updated the model implemented for the analysis, considering the most recent evidence.2The model was adapted considering the transitions among antiretroviral therapies (ARTs) observed in the trial, as reported in Figure 1, for years 1 and 2, and maintaining for years 3, 4, and 5 the same differential effectiveness (percentage of patients without virologic failure) observed between 48 and 96 weeks. In detail, the percentage of virologic failures considered in year 1 were 4.51% for ATV+r+2 NRTI and 0.76% for ATV+r+3TC; and in each following year were 8.66% for ATV+r+2 NRTI and 3.05% for ATV+r+3TC
What Should Guide Decision-Making Process in Urgency? A Curious Case of Right Acute Hemiparesis
No full text
Projection by Convolution: Optimal Sample Complexity for Reinforcement Learning in Continuous-Space MDPs
No full textCombined Multiportal Endoscopic Endonsasal and Transcrinial Approach for Recurnent Tuberculum SeIlae Meningioma: Operative Video
No full textSurgery for anterior cranial base lesions is challenging because tumors often spread from their origin to sinuses, orbits, and middle cranial fossa, resulting in risky surgeries.1,2 To approach such complex diseases, recently combined multiportal approaches have been proposed.3,4 At the best of our knowledge, operative application of a combined endoscopic endonasal and transcranial surgery for complex anterior cranial base lesions has not been described. Therefore a surgical video of such an approach is presented. A 37-year-old woman affected by a huge recurrence of a tuberculum meningioma extended to right orbit presented to our department, complaining of right ocular bulb dislocation with multidirectional limitations in eye movements (Video 1). Given the complexity and localization of the lesion, a combined endoscopic endonasal–transcranial surgery was performed. Such a combined approach, although demanding in terms of the presence of both otolaryngologist and neurosurgeon during the surgery, also requires strong synergy among them and permits them to control anterior cranial base lesions from both specialists' perspectives, simultaneously. In this case, while an endonasal corridor permitted an accurate excision of ethmoidal and medial orbital wall part of the lesion, a bicoronal approach allowed the aggression of the anterior cranial base portion of the tumor, allowing good control of cranial base neurovascular structures, eventually obtaining a gross total resection, without perioperative complications. Furthermore, a combined multiportal approach allows cooperative strategies among the surgeons involved, leading to safer, quicker, and more effective resections with less brain retraction, given the wide angles of views to the lesion that a multiportal approach can offer
Economic burden of the management of metastatic castrate-resistant prostate cancer in Italy: a cost of illness study
No full textUmberto Restelli,1,2 Giovanni Luca Ceresoli,3 Davide Croce,1,2 Laura Evangelista,4 Lorenzo Stefano Maffioli,5 Letizia Gianoncelli,3 Emilio Bombardieri6 1Center for Health Economics, Social and Health Care Management, LIUC – Università Cattaneo, Castellanza, Italy; 2School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa; 3Medical Oncology Department, Thoracic and Urologic Oncology Unit, Cliniche Humanitas Gavazzeni, Bergamo, 4Nuclear Medicine and Molecular Imaging Unit, Veneto Institute of Oncology IOV – IRCCS, Padova, 5Nuclear Medicine Department, Cancer Centre – ASST Ovest Milanese, Legnano, 6Nuclear Medicine Department, Cliniche Humanitas Gavazzeni, Bergamo, Italy Background: Prostate cancer (PCa) accounts for 20% of all cancers in subjects over 50 years in Italy. The majority of patients with PCa present with localized disease at the time of diagnosis, but many patients develop recurrent metastatic disease after treatment with curative intent. Androgen deprivation therapy is the standard of care for metastatic PCa patients; unfortunately, most of them progress to castrate-resistant prostate cancer (CRPC) within 5 years. Metastatic CRPC (mCRPC) heavily affects patients in terms of quality of life, side effects, and survival, and greatly impacts economic costs. The approval of new effective agents in recent years, including cabazitaxel, abiraterone acetate, enzalutamide, and radium-223, has dramatically changed patient management.Materials and methods: Here, we aimed to estimate the current costs of illness of mCRPC in Italy. All patients affected by mCRPC and treated with a single agent in an annual time horizon were considered. Therefore, the analysis was not focused on the management pathway of single patients through different lines of treatment. Direct medical costs referred to therapy, adverse event management, and skeletal-related event management were analyzed. A bottom-up approach was used to estimate the resource consumption: through national guidelines and expert opinions, the mean cost per patient was estimated and then multiplied by the total number of patients diagnosed with mCRPC. Results: Direct medical costs ranged from €196.5 million to €228.0 million, representing ~0.2% of the financing of the Italian National Health Service in 2016. The main cost driver was the cost of treatment, which represented more than 77% of the overall economic burden.Conclusion: Our analysis, reflective of real clinical practice, shows for the first time the high economic cost of mCRPC in Italy. Keywords: metastatic castrate-resistant prostate cancer, cost analysis, resource consumption, Italy, skeletal related events, oncolog
- …
