Table 3 Comparison of computational complexity and performance.

From: A cognitive internet of things resource allocation method based on multi-agent reinforcement learning algorithm

Algorithm

Avg. Training Time

Inference FLOPs

Avg. Inference Latency

IMAPPO (ours)

5.2 GPU-hours

1.5 GFLOPs

2.1 ms/decision

P-DDPG

4.5 GPU-hours

1.4 GFLOPs

2.0 ms/decision

MADDQN

3.8 GPU-hours

1.2 GFLOPs

1.9 ms/decision

MADQN

3.5 GPU-hours

1.2 GFLOPs

1.9 ms/decision

RANDOM

N/A

Negligible

0.1 ms (Negligible)

Back to article page