1: \begin{abstract}

2: We propose a novel deep neural network, coined DeepFPC-{$\ell_2$}, for solving the 1-bit compressed sensing problem. The network is designed by unfolding the iterations of the fixed-point continuation (FPC) algorithm with one-sided $\ell_2$-norm (FPC-$\ell_2$). The DeepFPC-$\ell_2$ method shows higher signal reconstruction accuracy and convergence speed than the traditional FPC-$\ell_2$ algorithm. Furthermore, we compare its robustness to noise with the previously proposed DeepFPC network---which stemmed from unfolding the FPC-$\ell_1$ algorithm---for different signal to noise ratio (SNR) and sign-flipped ratio (flip ratio) scenarios. We show that the proposed network has better noise immunity than the previous DeepFPC method. This result indicates that the robustness of a deep-unfolded neural network is related with that of the algorithm it stems from.

3: 

4: %DeepFPC has been proposed to show better performance than iterative one-sided \textit{$l_1$}-norm (FPC-\textit{$l_1$}) for solving the 1-bit compressed sensing problem. However, DeepFPC is less robust against noise, hence a robust algorithm is desired. In this paper, we propose an improved network,---DeepFPC-\textit{$l_2$}---, which is stemmed from  FPC-\textit{$l_2$} algorithm, and compare its robustness to noise with DeepFPC in different signal to noise ratio (SNR) and sign-flipped ratio (flip ratio) scenarios. We demonstrate that DeepFPC-\textit{$l_2$} offers higher signal reconstruction accuracy and speed than FPC-\textit{$l_2$} algorithm, and this network performs better compared to the DeepFPC when Gaussian noise or sign flippings exist in the measurements. The results also indicate that the robust property of these two unfolding network relies on the algorithm them stem from.

5: \end{abstract}

6: