core network, and only some infrastructure is performing the transmission. In fog side,
the number of terminals is many times compared to other two sides, and all IoT
terminals, multimedia laptops, Internet of Vehicles (IoV), etc., are connected among
each other and the eNodeBs, which will bring significant pressure to the whole
network.
From the discussion above, we can infer that the difficulties to evaluate the largescale complex network could be summarized as follows:
1. Number of nodes are huge, which makes the difficulty of summarizing the
parameters and decides which parameters are important [18].
2. Data transmitted through the line is also complex, and it is difficult to discover
which data is important.
3. The connection of network consists of multiple standard, and the fusion between
different types of communication will also bring delay, fault, etc. How to evaluate
the accuracy and the robust of data is still an open hard problem.
3 Generation of Deep Learning Network for Ranking
To tackle the evaluation of complex large-scale networks, we propose a novel deep
learning-based network parameter discovery and ranking scheme using the neural
network.
Content
Server
Content
Server
Wireless Server
Wireless Server
Cellular
IoT AP
ZigBee AP
BT AP
UE1
UE3
UE6
UE7
Ethernet
Content
Server
Wireless
Core Network
X2
X2
UE2
UE4
UE5
Content
Server
Content
Server
Cloud
Edge
Fog
Cellular
Cellular
Fig. 1. Structure of complex wireless networks. cloud–edge–fog structure is given according to
5G network.
548
Y. Gao et al.