Update paper.md

paper/paper.md

@@ -43,6 +43,8 @@ In this paper, we present OptiCommPy, an open-source Python package designed for
 
 The module structure of the OptiCommPy package is illustrated in Fig.~\ref{fig:pckg-struct}. At the top level, the package is named `optic`, containing five sub-packages: `comm, models, dsp, utils`, and `plot`.
 
+![Software description](OptiCommPy.png)
+
 The `comm` sub-package comprises three modules designed for implementing various digital modulation and demodulation schemes [@Proakis2001], including pulse amplitude modulation (PAM), quadrature amplitude modulation (QAM), phase-shift keying (PSK), and on-off keying (OOK). Evaluating the performance of these diverse digital communication schemes is made possible through different metrics, such as bit-error-rate (BER), symbol-error-rate (SER), error vector magnitude (EVM), mutual information (MI), and generalized mutual information (GMI) [@Alvarado2018], all available within the `comm.metrics` module.
 
 The `models` sub-package contains the majority of the mathematical/physical models used to build OptiCommPy simulations. Within the `models.devices` module, one can access models for a range of optical devices, encompassing optical Mach-Zehnder modulators, photodiodes, optical hybrids, optical coherent receivers, and more. These functions serve as fundamental building blocks for constructing simulations of optical transmitters and receivers. In the `models.channels` module, a collection of mathematical models for the fiber optic channel is provided, spanning from basic additive white Gaussian noise (AWGN) and linear propagation models to more sophisticated non-linear propagation models rooted in variants of the split-step Fourier method (SSFM)[@Agrawal2002]. In particular, it includes an implementation of the Manakov model to simulate nonlinear transmission over a fiber optic channel with polarization-multiplexing [@Marcuse1997a]. Certain computationally intensive models, such as the Manakov SSFM, have a CuPy-based version [@cupy_learningsys2017] accessible via `models.modelsGPU`, designed specifically for execution with CUDA GPU acceleration [@cuda].