Table 4 Learning and algorithm hyperparameters.

From: Adaptive and intelligent customized deep Q-network for energy-efficient task offloading in mobile edge computing environments

Parameter

Value

Discount factor (\(\gamma\))

0.9

Energy coefficient (\(\kappa\))

\(10^{-28}\)

Learning rate

0.001

Batch size

16

Optimizer

RMSProp

Exploration rate (\(\epsilon\))

1 \(\rightarrow\) 0.01 (decay)

Replay buffer

Per-device

Episodes

1000

Time slots per episode

100

Back to article page