As　a　promising　paradigm,　mobile　edge　computing　(MEC)　provides　cloud　resources　in　a　network　edge　to　offer　low-latency　services　to　mobile　devices　(MDs).　MEC　addresses　the　limited　resource　and　energy　issues　of　MDs　by　deploying　edge　servers,　which　are　often　located　in　small　base　stations.　It　is　a　big　challenge,　however,　as　how　to　dynamically　connect　resource-limited　MDs　to　nearby　edge　servers,　and　reduce　total　energy　consumption　by　MDs,　small　base　stations　and　a　cloud　data　center　(CDC)　all　in　a　hybrid　system.　To　tackle　the　challenge,　this　work　provides　an　intelligent　computation　offloading　method　for　both　static　and　dynamic　applications　among　entities　in　such　a　hybrid　system.　The　minimization　problem　of　total　energy　consumption　is　first　formulated　as　a　typical　mixed　integer　non-linear　program.　An　improved　meta-heuristic　optimization　algorithm,　named　Particle　swarm　optimization　based　on　Genetic　Learning　(PGL),　is　tailored　to　solve　the　problem.　PGL　synergistically　take　advantage　of　both　the　fast　convergence　of　particle　swarm　optimization,　and　the　global　search　ability　of　genetic　algorithm.　It　jointly　optimizes　task　offloading　of　heterogeneous　applications,　bandwidth　allocation　of　wireless　channels,　MDs＇　association　with　small　base　stations　and/or　a　cloud　datacenter,　and　computing　resource　allocation　of　MDs.　Numerical　results　with　real-life　system　configurations　prove　that　PGL　outperforms　several　state-of-the-art　peers　in　terms　of　total　energy　consumption　of　the　hybrid　system.　IEEE