Reinforcement Learning for Bit-Flipping Decoding of Polar Codes

. 2021 Jan 30;23(2):171. doi: 10.3390/e23020171

Algorithm 3 Q-learning-assisted Successive Cancellation Flip Decoding Algorithm

Input:

The received LLR:

y_{1}^{N}

Information bits length:

K

SNR:

s n r

The processed Q-table:

Q

Output: The decoded vector:

{\hat{u}}_{1}^{N}

Initialization:

Q_{f l i p} \leftarrow g e t f l i p s e t (s n r, Q)

;

{\hat{u}}_{1}^{N} \leftarrow Q L S C F (y_{1}^{N}, K, Q_{f l i p})

;
Return

{\hat{u}}_{1}^{N}

;
// The QLSCF function
Function

{\hat{u}}_{1}^{N} \leftarrow Q L S C F (y_{1}^{N}, K, Q_{f l i p})

:
For

i \leftarrow 1

| Q_{f l i p} |

a c t i o n \leftarrow Q_{f l i p} (i)

;

{\hat{u}}_{1}^{N} \leftarrow S C F d e c o d e r (y_{1}^{N}, K, a c t i o n)

;
If

C R C c h e c k ({\hat{u}}_{1}^{N})

successes:

b r e a k

;
End if
End for
Return

{\hat{u}}_{1}^{N}

;
End function