Increasing the functional unit numbers, density, and diversity on chip has long been a goal in the development of integrated photonics. Major progress came from the development of III-V integrated photonics for communication applications. However, Si photonics has seen much progress and promises new applications. Hybrid approaches further complicate the development of VLSIP and scaling methods from a foundry and supply chain perspective.
At the same time new applications are emerging in the areas of IoT and 5G wireless that are demanding more from optical devices and systems in cost and energy sensitive regimes that require integration. The first half of this workshop will examine the current state of large-scale integration across multiple integrated photonic platforms and examine the current and emerging application drivers, including what future applications are motivating larger scale integration and which have the potential market size to support VLSIP. It will then explore the challenges to realizing VLSIP related to fabrication, foundry support, and supply chains across the multiple technology and integration platforms.
Hybrid Photonic Integrated Circuits (Hybrid PICs) are complex and cost-effective at the same time. This half-day workshop will cover all topics of Hybrid PICs from materials to integration technologies, enabling platforms and modelling tools towards their use in different fields of applications such as communications, quantum technologies, analytics, and sensing.
Neuromorphic computing is a signal processing technology which mimics neuro-biological architectures in the brain. Neuromorphic signal processing has potential in the applications of generation, equalization or detection of signals with the perspective benefit of a significantly higher power efficiency than today's digital signal processing (DSP) technology. Neuromorphic computing or processing approaches are based on analog signal processing with electrical or optical signal, partly digital or asynchronous processing, ranging from emulation of spiking neuron to analog computation of artificial neural network and reservoir computing.
New generation of brain-inspired signal processing algorithms as well as recent advances in development of new materials and novel architectures enabled a surge in optical implementation of machine learning methods. This has become a rapidly growing area. Recent achievements include, for instance, demonstrations of optical neural networks implementing speech recognition and mitigation of distortions in optical communications. Different technological avenues are being pursued, including bulk and integrated optics, while several start-ups have been recently created to exploit these opportunities.
The aim of this workshop is to bring together leading experts in photonics, optical communications and neuromorphic engineering to examine the future of optical neuromorphic computing.
The workshop will cover specifically the following main topics:
1. Scaling of neuromorphic technology for high-performance signal processing
2. Applications for neuromorphic engineering
3. Low-complexity solutions for short-reach applications
Key questions which will be addressed during the workshop are:
a. What is neuromorphic computing and how does it differ from machine learning?
b. What are the advantages / disadvantages of spiking versus continuous neuromorphic approaches?
c. Can neuromorphic signal processing reduce power consumption with respect to digital electronic signal processing? By how much?
d. Is neuromorphic signal processing compatible with the speed requirements of optical communications?
e. What are the challenges in digital signal processing that could be addressed by a neuromorphic approach?
f. What are the simplest signal processing tasks for which a neuromorphic approach could prove beneficial?
Efforts in enabling higher capacity tend to rely on constellation energy efficiency, multi-mode and multi-core fibres. But isn't it time to truly explore other transmission windows? For standard single mode fibre, work is emerging on the efficient use of S & O bands for short reach communications. Alternatively, hollow core fibres have the advantage of significant lower nonlinear interaction, benefiting for high power operation, and potentially shifting transmission to 2 µm. This workshop aims to have a focus panel discussion that follows a series of short talks from stakeholders in industry and academia, with the aim to draft the challenges and needs to open these new transmission windows.
Modeling has been the cornerstone of advances in fiber systems, from understanding our devices to untangling transmission effects at any and all distances. The discoveries that push communications speeds and reach, also challenge our ability to understand and anticipate the implications of adopting each new technique.
We will explore in this workshop, both the changes driving the need for new models, and the disruptive application of machine learning to supplant even the idea of modeling. Transmission systems spanning 100s of nm, exploiting dozens of spatial modes in exotic fibers, with DSP that unknots correlations in nonlinear effects, and all at transoceanic distances are under investigation. These are the advances that challenge models for fiber transmission.
How do we capture Raman effects in UWB transmissions, acoustic Brouillon scattering in undersea cables, quantum capacity limits, and more? Device designs and system architectures must be optimized to maximize capacity.
Adopting silicon photonics brings low cost, but optical components behave differently when born on a CMOS platform. SOAs are old friends, but how will they react when we push them to cover UWB? Should we model these devices & systems, or should we let a machine help us to maximize capacity instead? We will bring together experts to share their views on emerging challenges in modeling as a tool to ever improved optical communications.
Topics: Very high capacity / low latency optical fronthaul; Next generation fronthaul protocols suitable for optical, low latency and high bandwidth communication; Self-learning and programable optical network for convergence; Role of optical network in disaggregation , function splicing and slicing; Quantum secured network infrastructure and functions; Role of optical Data centre in ultra-low latency and highly mobile applications and services; Role of optical network in cloudification of network function; 5G testbed implementations and optical fronthauls deployment including real-world lessons learned
The workshop will also include a panel discussion at the end.
The next line rate for PON has been agreed in the ITU-T as 50G. This workshop will aim to address the requirements/applications for such capacity in PON and the relevant technologies to achieve 50G+ at the low costs required for access applications.
Novel applications including but not restricted to 5G and the (industrial) Internet of Things – (I)IoT – rely on low latency communication. This workshop will investigate how optical networks can enable those applications.