Graph signal sampling via reinforcement learning

 |  Login

Show simple item record

dc.contributor Aalto-yliopisto fi
dc.contributor Aalto University en
dc.contributor.advisor Jung, Alexander Abramenko, Oleksii 2018-11-13T13:39:11Z 2018-11-13T13:39:11Z 2018-11-07
dc.description.abstract Graph signal sampling is one the major problems in graph signal processing and arises in a variety of practical applications, such as data compression, image denoising and social network analysis. In this thesis we formulate graph signal sampling as a reinforcement learning (RL) problem, which unleashes powerful methods developed recently within RL. Within our approach the signal sampling is carried out by an agent which crawls over the empirical graph and selects the most relevant graph nodes to sample, i.e., determines the corresponding graph signal values. Overall, the goal of the agent is to select signal samples which allow for the smallest graph signal recovery error. The behavior of the sampling agent is described using a policy which determines whether or not a particular node should be sampled. The policy is continuously adjusted which implies an inherent robustness to changes in the data structure. We propose two RL-based sampling algorithms and evaluate their performance by means of illustrative numerical experiments. After this we conduct elaborative discussion on strengths and weaknesses of the proposed solutions and explain phenomenons observed during simulations. Based on the simulation results, we identify some major challenges related to application of RL to graph signal sampling and propose possible solutions leading to prospects for the future work. en
dc.format.extent 59
dc.format.mimetype application/pdf en
dc.language.iso en en
dc.title Graph signal sampling via reinforcement learning en
dc.type G2 Pro gradu, diplomityö fi Perustieteiden korkeakoulu fi
dc.subject.keyword reinforcement learning en
dc.subject.keyword graph signal sampling en
dc.subject.keyword convex optimization en
dc.subject.keyword multi-armed bandit en
dc.subject.keyword policy gradient en
dc.subject.keyword complex networks en
dc.identifier.urn URN:NBN:fi:aalto-201811135787
dc.programme.major Machine Learning and Data Mining fi
dc.programme.mcode SCI3044 fi
dc.type.ontasot Master's thesis en
dc.type.ontasot Diplomityö fi
dc.contributor.supervisor Jung, Alexander
dc.programme Master’s Programme in Computer, Communication and Information Sciences fi
local.aalto.electroniconly yes
local.aalto.openaccess yes

Files in this item

This item appears in the following Collection(s)

Show simple item record

Search archive

Advanced Search

article-iconSubmit a publication


My Account