The article discusses the challenges faced by developers in Deep Reinforcement Learning (DRL) applications. It presents a taxonomy of challenges based on a large-scale empirical study of Stack Overflow posts. The study categorizes challenges into DRL issues, DL issues, DRL libraries/frameworks, parallel processing & multi-threading, and general programming issues. Key insights include the prevalence of challenges like comprehension, API usage, and design problems. The survey validation confirms that practitioners encounter these challenges and perceive them as critical with medium to high effort required for resolution.
toiselle kielelle
lähdeaineistosta
arxiv.org
Tärkeimmät oivallukset
by Mohammad Meh... klo arxiv.org 03-22-2024
https://arxiv.org/pdf/2310.09575.pdfSyvällisempiä Kysymyksiä