Abstract

Generative AI-based applications accelerate the technological evolution of computing, memory, and broadband interconnect. Currently, interconnect bandwidth evolution lags behind the computing counterpart. In this context, high-speed optical interconnects for chip-to-chip, chip-tomodule and module-to-module urgently needs scale-up. The core requirement for such applications is to achieve the highest possible capacity density and power efficiency. Coherent optical communication is expected to penetrate into this short-reach scenario in future when 1.6T and 3.2T optical module is considered. Coherent DSP plays a central role in determining performance and power consumption. In this talk, a new simplified coding and DSP algorithm based on the standard coherent optical transceiver will be proposed, including extremely low-complexity probability constellation shaping (PCS), timing recovery(TR) and multiple input and fractional-spaced multiple output (MIMO) equalization. Technical feasibility of the proposed low power DSP algorithms is verified via experimental offline and/or FPGA demonstrations.

Abstract

The QBIC observed in metasurfaces provides a feasible platform to achieve in-plane strong light-matter interaction, as well as to develop ultrasensitive biosensor. However, the existing metasurface designs are difficult to realize highly efficient excitation and high-performance sensing of QBIC in terahertz (THz) band. Here, we manipulate the interference coupling between electric quadrupole and magnetic dipole by introducing an asymmetry α into the metallic metasurface structure, which excites ultrahigh quality QBIC resonance with Q factor of up to 503. Correspondingly, light field energy constrained by the metasurface and effective sensing area achieved enormous increases of about 400% and 1300%, respectively, which greatly expands the spatial extent and intensity of light-matter interaction. Simulations and experiments show that the proposed QBIC metasurface deliver a high refractive index sensitivity reaching 420 GHz/RIU, where RIU is the refractive index unit, and its direct limit of detection (LoD) for trace homocysteine (Hcy) molecules is 12.5 pmol/μL. Its performance is about 40-times better than that of the classical Dipole mode.

Abstract

Introduction: There are several surgical options described for the treatment of anorectal fistulas, specially in complex cases where recurrence rates and the possibility of postoperative complications are still high. FILAC has been described in this study as an option in the management of anorectal fistula.

Objective: The aim of this study was to assess the novel radial laser probe treatment in complex fistula in-ano and report the success rate and recurrence rate.

Methods: We studied retrospectively 56 patients who, according to our hospital patient records, underwent radial laser probe surgery between March 2019 and August 2020.

Results: In a mean follow-up time of 6 months, the success rate at 2 months was 86.0%. Most operations were done under spinal anesthesia. The recovery time was rapid, and median sick leave was 7 days. Of those initially successfully treated, 3.0% developed a recurrence. Altogether 4.0% of the patients underwent a re-operation.

Conclusion: There is a good success rate using FiLaC™ treatment. FiLaC™ is very effective in treatment of complex fistula and as well as recurrence of fistula. It has a short hospital stay and as well as it is painless surgical technique that should be largely used in our country.

Keywords: Anal fistula, FiLaC, Sphincter-preserving, LASER

The purpose of this work is to provide a discussion into the use of deep reinforcement learning techniques in elastic optical networks (EONs) to properly tackle the problem of routing, modulation, and spectrum assignment. By providing an insight into current optical networks, we are able to understand the current underlying problem of these networks and how EONs provide a proper solution with quite a few benefits. However, this innovative approach has its fair share of problems that need to be addressed before implementing this new system in a commercial setting. By trying to resolve the issues and comparing them to currently existing solutions, one can properly assess the benefits and choose to move forward with the one that best suits the current needs of the market. This research intends to address this problem by employing a hybrid machine learning (ML) based approach. The proposed approach aims to solve the routing and spectrum assignment (RSA) problem by employing a fragmentation aware RSA (FA-RSA) algorithm . The proposed algorithm identifies an optimal route for the network traffic and allocates a specific number of spectrum slots to maximize the spectrum utilization. The proposed approach aims to achieve an appropriate balance between the fault restoration and blocking probability. Results validate the efficacy of the proposed. The proximal policy optimization (PPO) algorithm is used to optimize the system. PPO algorithm used to generate the data needed to properly interpret random behaviors serves as a baseline for what sort of improvements can be expected. Furthermore, properly analyzing this data and providing a proper medium to visualize and digest this data allows the consumer to get a better understanding of what benefits will arise from this new approach to a current system. The system showed a better adaptation of considerably lower blocking probability at heavier loads which would be the real life high-trafficked scenario. The proposed approach is also evaluated in terms of queue length, average waiting time, number of spectrum fragments, overall blocking probability, reward rate, average utilization ratio and fault restoration ratio. The research also gives a probabilistic approach for effective resource utilization thereby improving the efficiency of the network.