In this paper, we describe the design of Neurogrid, a neuromorphic system for simulating large-scale neural models in real time. Neuromorphic systems realize the function of biological neural systems by emulating their structure. Designers of such systems face three major design choices: 1) whether to emulate the four neural elements-axonal arbor, synapse, dendritic tree, and soma-with dedicated or shared electronic circuits; 2) whether to implement these electronic circuits in an analog or digital manner; and 3) whether to interconnect arrays of these silicon neurons with a mesh or a tree network. The choices we made were: 1) we emulated all neural elements except the soma with shared electronic circuits; this choice maximized the number of synaptic connections; 2) we realized all electronic circuits except those for axonal arbors in an analog manner; this choice maximized energy efficiency; and 3) we interconnected neural arrays in a tree network; this choice maximized throughput. These three choices made it possible to simulate a million neurons with billions of synaptic connections in real time-for the first time-using 16 Neurocores integrated on a board that consumes three watts.

/pdf/neurogrid-a-mixed-analog-digital-multichip-system-for-large-494ipl0j1g.pdf

Neurogrid: A Mixed-Analog-Digital Multichip System for Large-Scale Neural Simulations

Neuromorphic computing has come to refer to a variety of brain-inspired computers, devices, and models that contrast the pervasive von Neumann computer architecture This biologically inspired approach has created highly connected synthetic neurons and synapses that can be used to model neuroscience theories as well as solve challenging machine learning problems The promise of the technology is to create a brain-like ability to learn and adapt, but the technical challenges are significant, starting with an accurate neuroscience model of how the brain works, to finding materials and engineering breakthroughs to build devices to support these models, to creating a programming framework so the systems can learn, to creating applications with brain-like capabilities In this work, we provide a comprehensive survey of the research and motivations for neuromorphic computing over its history We begin with a 35-year review of the motivations and drivers of neuromorphic computing, then look at the major research areas of the field, which we define as neuro-inspired models, algorithms and learning approaches, hardware and devices, supporting systems, and finally applications We conclude with a broad discussion on the major research topics that need to be addressed in the coming years to see the promise of neuromorphic computing fulfilled The goals of this work are to provide an exhaustive review of the research conducted in neuromorphic computing since the inception of the term, and to motivate further work by illuminating gaps in the field where new research is needed

A Survey of Neuromorphic Computing and Neural Networks in Hardware.

Synapses are crucial elements for computation and information transfer in both real and artificial neural systems. Recent experimental findings and theoretical models of pulse-based neural networks suggest that synaptic dynamics can play a crucial role for learning neural codes and encoding spatiotemporal spike patterns. Within the context of hardware implementations of pulse-based neural networks, several analog VLSI circuits modeling synaptic functionality have been proposed. We present an overview of previously proposed circuits and describe a novel analog VLSI synaptic circuit suitable for integration in large VLSI spike-based neural systems. The circuit proposed is based on a computational model that fits the real postsynaptic currents with exponentials. We present experimental data showing how the circuit exhibits realistic dynamics and show how it can be connected to additional modules for implementing a wide range of synaptic properties.

/pdf/synaptic-dynamics-in-analog-vlsi-2pbykkhvp1.pdf

Synaptic Dynamics in Analog VLSI

A great deal of work has been done to analyze the problem of robot move sequencing and part scheduling in robotic flowshop cells. We examine the recent developments in this literature. A robotic flowshop cell consists of a number of processing stages served by one or more robots. Each stage has one or more machines that perform that stage's processing. Types of robotic cells are differentiated from one another by certain characteristics, including robot type, robot travel-time, number of robots, types of parts processed, and use of parallel machines within stages. We focus on cyclic production of parts. A cycle is specified by a repeatable sequence of robot moves designed to transfer a set of parts between the machines for their processing. We start by providing a classification scheme for robotic cell scheduling problems that is based on three characteristics: machine environment, processing restrictions, and objective function, and discuss the influence of these characteristics on the methods of analysis employed. In addition to reporting recent results on classical robotic cell scheduling problems, we include results on robotic cells with advanced features such as dual gripper robots, parallel machines, and multiple robots. Next, we examine implementation issues that have been addressed in the practice-oriented literature and detail the optimal policies to use under various combinations of conditions. We conclude by describing some important open problems in the field.

https://papers.ssrn.com/sol3/Delivery.cfm/SSRN_ID1295004_code328623.pdf?abstractid=1128751&mirid=1&type=2

Sequencing and Scheduling in Robotic Cells: Recent Developments

A great deal of work has been done to analyze the problem of robot move sequencing and part scheduling in robotic flowshop cells. We examine the recent developments in this literature. A robotic flowshop cell consists of a number of processing stages served by one or more robots. Each stage has one or more machines that perform that stage's processing. Types of robotic cells are differentiated from one another by certain characteristics, including robot type, robot travel-time, number of robots, types of parts processed, and use of parallel machines within stages. We focus on cyclic production of parts. A cycle is specified by a repeatable sequence of robot moves designed to transfer a set of parts between the machines for their processing.

We start by providing a classification scheme for robotic cell scheduling problems that is based on three characteristics: machine environment, processing restrictions, and objective function, and discuss the influence of these characteristics on the methods of analysis employed. In addition to reporting recent results on classical robotic cell scheduling problems, we include results on robotic cells with advanced features such as dual gripper robots, parallel machines, and multiple robots. Next, we examine implementation issues that have been addressed in the practice-oriented literature and detail the optimal policies to use under various combinations of conditions. We conclude by describing some important open problems in the field.

/pdf/sequencing-and-scheduling-in-robotic-cells-recent-554zpql4t3.pdf

Cluster tools are highly integrated machines that can perform a sequence of semiconductor manufacturing processes. Their integrated nature can complicate analysis when evaluating how process changes affect the overall tool performance. This paper presents two integrated models for understanding the behavior of a simple, single loadlock cluster tool. The first model is a network model that evaluates the total lot processing time for a given sequence of activities. By including a manufacturing process model (in the form of a response surface model, or RSM), the model calculates the lot makespan, the total time to process a lot of wafers, as a function of the process parameter values and other operation times. This model allows us to quantify the sensitivity of total lot processing time with respect to process parameters and times. In addition, we present an integrated simulation model that includes a process model. For a given scheduling rule that the cluster tool uses to sequence wafer movements, we can use the simulation to evaluate the impact of process changes, including changes to product characteristics and changes to process parameter values. In addition, we can construct an integrated network model to quantify the sensitivity of total lot processing time with respect to process times and process parameters in a specific scenario. We also present an evaluation of the effectiveness of two different scheduling rules, push and pull. The examples presented here illustrate the types of insights that we can gain from using such methods. Namely, the lot makespan is a function not simply of each operation's process time, but specifically of the chosen process parameter values. Modifying the process parameter values may also have significant impacts on the manufacturing system performance, a consequence of importance that is not readily obvious to a process engineer when tuning a process. This result can be seen either with the decrease of raw process time causing little change to the makespan, or the extreme example in which this could cause an increase in makespan because of an inefficient scheduling rule. Additionally, because the cluster tool's maximum throughput, which is the inverse of the lot makespan, depends on the process parameters, the tradeoffs between process performance and throughput should be considered when evaluating potential process changes and their manufacturing impact.

/pdf/evaluating-the-impact-of-process-changes-on-cluster-tool-1vx01i0otq.pdf

Evaluating the impact of process changes on cluster tool performance

Synapses are a critical element of biologically-realistic, spike-based neural computation, serving the role of communication, computation, and modification. Many different circuit implementations of synapse function exist with different computational goals in mind. In this paper we describe a new CMOS synapse design that separately controls quiescent leak current, synaptic gain, and time-constant of decay. This circuit implements part of a commonly-used kinetic model of synaptic conductance. We show a theoretical analysis and experimental data for prototypes fabricated in a commercially-available 1.5µm CMOS process.

/pdf/a-summating-exponentially-decaying-cmos-synapse-for-spiking-3gcv9xvyc3.pdf

A Summating, Exponentially-Decaying CMOS Synapse for Spiking Neural Systems

Simulation models are useful to predict and understand the impact of changes to a manufacturing system. Typical factory simulation models include the parts being manufactured in the factory and the people and resources processing and handling the parts. However, these models do not include equipment or process details, which can affect operational performance such as cycle time and inventory. Separate models are used to evaluate processes and equipment. Thus, it is difficult to evaluate the operational impact of equipment or process changes. However, this information could help factory managers and manufacturing process engineers make better decisions when changing processes or selecting equipment configurations. This paper describes a heterogeneous simulation environment for understanding how equipment and process changes affect the performance of a wafer fabrication facility. This integrated tool incorporates response surface models that describe process behavior, operational and optimization models of equipment behavior, and a discrete event simulation model of factory operations. Thus, the tool can measure how process changes and equipment configuration changes change the system performance. We have applied this tool to a specific wafer fab problem.

/pdf/understanding-the-impact-of-equipment-and-process-changes-jf1wm2tfmx.pdf

Rock Z. Shi

Papers

Evaluating the impact of process changes on cluster tool performance

A Summating, Exponentially-Decaying CMOS Synapse for Spiking Neural Systems

Understanding the impact of equipment and process changes with a heterogeneous semiconductor manufacturing simulation environment