Optimizing Image Classification: Automated Deep Learning Architecture Crafting with Network and Learning Hyperparameter Tuning

. 2023 Nov 4;8(7):525. doi: 10.3390/biomimetics8070525

Algorithm 4: Competency-Based Learning in ETLBOCBL-CNN’s Modified Teacher Phase
Inputs: $P = [X_{1}, \dots, X_{n}, \dots ., X_{N}]$ , $N$ , $D$ , $X^{T e a c h e r}$ , $ℛ^{t r a i n}$ , $ℛ^{v a l i d}$ , $S^{b a t c h}$ , $ε^{t r a i n}$ , $R^{L}$ , $C^{n u m}$ , $S^{G r o u p}$ , $G$
01:	Initialize offspring population set as $P^{o f f} \leftarrow \emptyset$ ;
02:	Calculate the population mean $\bar{X} . M e a n$ using Algorithm 3;
03:	Sort all solution members of $P$ ascendingly by referring to their fitness values $X_{n} . E r r$ ;
04:	Determine the group index g assigned to $X_{n} . G r p$ of all sorted learners using Equations (9) and (10);
05:	for $n = 1$ to $N$ do
06:	Initialize the n-th offspring learner as $X_{n}^{o f f} \leftarrow \emptyset$ ;
07:	if $X_{n} . G r p \geq 3$ then
08:	Randomly select two better group indices of $g_{r 1}, g_{r 2} \in {1, g - 1},$ where $g_{r 1} < g_{r 2} < g;$
09:	Randomly select two predominant learners with the population indices represented as
	$n_{r 1}^{g_{r 1}} \in {(g_{r 1} - 1) S^{G r o u p}, g_{r 1} S^{G r o u p}}$ and $n_{r 2}^{g_{r 2}} \in {(g_{r 2} - 1) S^{G r o u p}, g_{r 2} S^{G r o u p}};$
10:	Calculate $X_{n}^{o f f} . P o s$ using Equation (11);
11:	else if $X_{n} . G r p = 2$ then
12:	Randomly select two predominant learners from the first group (i.e., $g = 1$ ) with the population
	indices of $n_{r 1}^{1}, n_{r 2}^{1} \in {1, S^{G r o u p}}$ and $n_{r 1}^{1} \neq n_{r 2}^{1},$ where $X_{n_{r 1}^{1}} . G r p = X_{n_{r 2}^{1}} . G r p = 1$ ;
13:	Compare the fitness values of two predominant learners, i.e., $X_{n_{r 1}^{1}} . E r r$ and $X_{n_{r 2}^{1}} . E r r$ ;
14:	Calculate $X_{n}^{o f f} . P o s$ using Equation (12);
15:	else if $X_{n} . G r p = 1$ then
16:	Calculate $X_{n}^{o f f} . P o s$ using Equation (13);
17:	end if
18:	for $d = 1$ to D do
19:	if $d \neq 2 N_{m a x}^{C o n v} + 3 l - 1$ with $l = 1, \dots, N_{m a x}^{C o n v}$ then
20:	$X_{n}^{o f f} . P o s_{d} \leftarrow R o u n d (X_{n}^{o f f} . P o s_{d})$ ;
21:	end if
22:	end for
23:	Perform fitness evaluation on $X_{n}^{o f f} . P o s$ to obtain $X_{n}^{o f f} . E r r$ using Algorithm 2;
24:	if $X_{n}^{o f f} . E r < X^{T e a c h e r} . E r r$ then
25:	$X^{T e a c h e r} . P o s \leftarrow X_{n}^{o f f} . P o s$ , $X^{T e a c h e r} . E r r \leftarrow X_{n}^{o f f} . E r r$
26:	end if
27:	$P^{o f f} \leftarrow P^{o f f} \cup X_{n}^{o f f}$ ;
28:	end for
Outputs: $P^{o f f} = [X_{1}^{o f f}, \dots, X_{n}^{o f f}, \dots ., X_{N}^{o f f}]$ , $P = [X_{1}, \dots, X_{n}, \dots ., X_{N}]$ , $X^{T e a c h e r}$