From: DQN-based resource allocation for NOMA-MEC-aided multi-source data stream
Notation | Definition |
---|---|
The number of source S | 5 |
The variance of the AWGN | 0.01 |
Wireless bandwidth B | 6Â MHz |
Local computing capability \(f_s\) | \(5 \times (10^7)\) cycle/s |
Maximum transmit power \(P_{\max }^s\) | 1W |
Local calculated power of the sth source \(P_{\rm local}^s\) | 1.5W |
The calculated power of the CAP \(P_{\rm MEC}^s\) | 1.5W |
The computing power allocated to each source of the CAP \(F_s\) | \(8 \times (10^7)\) cycle/s |
The weight factor of system cost \(\mu\) | 0.5 |
The loss function of neural network in DQN | \(10^{-4}\) |
Learning rates of neural network in DQN | \(10^{-2}\) |
The capability of experience replay | \(2 \times (10^4)\) |
The batch size of neural network in DQN | 64 |