Fog　radio　access　networks　(F-RANs)　are　regarded　as　a　promising　solution　for　the　future　wireless　networks　by　incorporating　the　concept　of　fog　computing　into　the　architecture　of　wireless　access　networks.　F-RANs　make　full　use　of　the　computing　and　storage　capabilities　of　edge　network　devices,　and　can　effectively　alleviate　the　overhead　of　backhaul　links　and　reduce　the　processing　delay　of　the　tasks.　However,　existing　researches　were　mostly　focused　on　the　coupled　uplink　and　downlink　scenario　which　will　lead　to　insufficient　utilization　of　resources　and　high　energy　consumption.　In　order　to　solve　above　problems,　we　introduce　the　decoupled　the　uplink　and　downlink　(DUDA)　access　mode　to　F-RANs　and　jointly　optimize　the　user　access　and　computation　offloading　decisions　with　the　aim　of　energy　consumption　minimization.　To　reduce　the　computational　complexity,　we　further　utilize　the　reinforcement　learning　algorithm　to　solve　this　complex　combinatorial　optimization　problem.　The　simulation　results　show　that　compared　with　other　schemes,　the　proposed　algorithm　can　significantly　improve　the　energy　consumption　performance　while　meeting　user　demands.　©　2019　IEEE.