Algorithm 1: Discriminative fine-tuning algorithm.
1:	Procedure DFT
2:		Input:				$({\alpha }_{min}:$ minimum learning rate, ${\alpha }_{max}$: maximum learning rate, $m_{min}$: minimum momentum, $m_{max}$: maximum momentum, $\mathcal{D}$: size of dataset, batch_size)
3:		Output:				$(\theta :$ Network parameters)
4:		$t \leftarrow \frac{\mathcal{D}}{batch\_size}$
5:		$\kappa \leftarrow rand\left(0,1\right)$ $//\kappa$    //κ determines how rapidly the learning rate increase or reduces
		while $\kappa t<\frac{1}{2}{t}_{max}$
6:			for $t$ in each iteration do:
7:				for $l$ in each layer do:
8:					${\alpha }_t^l\leftarrow {\alpha }_{min}+\left(\frac{{\alpha }_{max}-{\alpha }_{min}}{{\alpha }_{max}}\right)l$     //increase learning rate per layer
9:					$m_t^l\leftarrow m_{min}+\left(\frac{m_{max-m_{min}}}{m_{max}}\right)l$   //increasing the momentum per layer
10:					${𝓋}_{t+1}^l\leftarrow m_t^l𝓋{ }_t^l-{\alpha }_t^l\frac{dJ\left({\theta }_t^l\right)}{d{\theta }_t^l}$
11:					${\theta }_{t+1}^l\leftarrow {\theta }_t^l+{ }_{t+1}^l$             //update the layer parameters
12:					end for
13.				end for
14.			end while
15.		while $\frac{1}{2}t\le \kappa t<t_{max}$
16.			for t in each iteration do:
17.				for $l$ in each layer do:
18.					${\alpha }_t^l\leftarrow {\alpha }_{min}-\left(\frac{{\alpha }_{max}-{\alpha }_{min}}{{\alpha }_{max}}\right)l$    //increase learning rate per layer
19.					$m_t^l\leftarrow m_{min}-\left(\frac{m_{max-m_{min}}}{m_{max}}\right)l$    //increasing the momentum per layer
20.					${𝓋}_{t+1}^l\leftarrow m_t^l𝓋{ }_t^l-{\alpha }_t^l\frac{dJ\left({\theta }_t^l\right)}{d{\theta }_t^l}$
21.					${\theta }_{t+1}^l\leftarrow {\theta }_t^l+{ }_{t+1}^l$          //update the layer parameters
22.					end for
23.				end for
24.			end while
25.	end Procedure