English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Incomplete knowledge of the environment leads an agent to make decisionsunder uncertainty. One of the major dilemmas in Reinforcement Learning (RL)where an autonomous agent has to balance two contrasting needs in making itsdecisions is: exploiting the current knowledge of the environment to maximizethe cumulative reward as well as exploring actions that allow improving theknowledge of the environment, hopefully leading to higher reward values(exploration-exploitation trade-off). Concurrently, another relevant issueregards the full observability of the states, which may not be assumed in allapplications. Such as when only 2D images are considered as input in a RLapproach used for finding the optimal action within a 3D simulationenvironment. In this work, we address these issues by deploying and testingseveral techniques to balance exploration and exploitation trade-off onpartially observable systems for predicting steering wheels in autonomousdriving scenario. More precisely, the final aim is to investigate the effectsof using both stochastic and deterministic multi-armed bandit strategiescoupled with a Deep Recurrent Q-Network. Additionally, we adapted and evaluatedthe impact of an innovative method to improve the learning phase of theunderlying Convolutional Recurrent Neural Network. We aim to show that adaptivestochastic methods for exploration better approximate the trade-off betweenexploration and exploitation as, in general, Softmax and Max-Boltzmannstrategies are able to outperform epsilon-greedy techniques.

Impact of multi-armed bandit strategies on deep recurrent reinforcement learning