Computational Offloading in Mobile Edge with Comprehensive and Energy Efficient Cost Function: A Deep Learning Approach

. 2021 May 19;21(10):3523. doi: 10.3390/s21103523

Algorithm 1 Partial Offloading with Partitioning

Input: { $m, γ, n, r_{i}, k_{i}, λ_{i}$ }

Output: { $z^{*}, o p^{*}$ }

1:
$PR \leftarrow$ Matrix of possible partitions
2:
$OP \leftarrow$ Matrix of possible offloading policies
3:
for $k = 1 :^{M} Z_{i}$ do
4:
$P (k) \leftarrow PR (k, :)$
5:
for $l = 1 : 2^{i}$ do
6:
$e (l) \leftarrow OP (l, :)$
7:
for $j = 1 : i$ do
8:
if $(e (j) = 0)$ then
9:
$c o s t (j) \leftarrow f_{l c}$ using $(18)$
10:
else
11:
$c o s t (j) \leftarrow f_{o c}$ using $(19)$
12:
end if
13:
end for
14:
$c o s t_{1} (l) \leftarrow$ sum( $c o s t$ )
15:
end for
16:
$[i n d e x, c o s t_{2} (k)] \leftarrow m i n (c o s t_{1})$
17:
${OP}_{1} (k) \leftarrow OP (i n d e x, :)$
18:
end for
19:
$[i n d e x, c o s t_{m i n i m u m}] \leftarrow m i n (c o s t_{2})$
20:
$o p^{*} \leftarrow {OP}_{1} (i n d e x, :)$
21:
$z^{*} \leftarrow PR (i n d e x, :)$
22:
Save input data: $i n p u t \leftarrow {m, γ, n, r_{i}, k_{i}, λ_{i}}$
23:
Save output data: $l a b e l s \leftarrow {z^{*}, o p^{*}}$
24:
Train the DNN: $T r a i n e d_D N N \leftarrow t r a i n (i n p u t, l a b e l s)$