Explore the vast range of titles available.

Custom integrated circuits modeling biological neural networks serve as tools for studying brain computation and platforms for exploring new architectures and learning rules of artificial neural networks. Time synchronization across network units is an important aspect of these designs to ensure reproducible results and maintain hardware-software equivalence. Current approaches rely on global synchronization protocols, which fundamentally limit system scalability. To overcome this, we develop NeuroScale, a decentralized and scalable neuromorphic architecture that uses local, aperiodic synchronization to preserve determinism without global coordination. Cores of co-localized compute and memory elements model neural and synaptic processes, including spike filtering operations, subthreshold neural dynamics, and online Hebbian learning rules. Multiple cores communicate via spikes across a routing mesh, using distributed event-driven synchronization to efficiently scale to large networks. We compare this synchronization protocol to the global barrier synchronization approaches of IBM TrueNorth and Intel Loihi, demonstrating NeuroScale’s advantages for large system sizes. Current deterministic neuromorphic systems rely on global barrier synchronization to achieve reproducible results which limits their scalability as system size grows. Here, the authors present NeuroScale, a deterministic and scalable architecture that replaces global synchronization with distributed local synchronization.

In conventional frame-based image sensors, every pixel records brightness information and sends this information to a receiver serially in a scanning fashion. This full-frame readout approach suffers from high bandwidth requirements and increased power consumption with the increasing size of the pixel array. Event-based image sensors are gaining popularity for reducing the bandwidth and power requirements by sending only meaningful data in an event-driven approach with the help of address-event representation (AER) communication protocol. However, the event-based readout suffers from increased latency and timing error when the number of pixels with an event increase. In this paper, we introduce a new field-programmable AER (FP-AER) encoding scheme which offers benefits of both frame-based and event-based approaches. The readout design can be configured \"in the field\" using configuration bits. We also compare the performance of the proposed design against existing AER-based approaches for imaging applications and show that FP-AER performs best in both scanning and event-based readout.

Contact tracing is a vital public health tool used to prevent the spread of infectious diseases. However, traditional interview-format contact tracing (TCT) is labor-intensive and time-consuming and may be unsustainable for large-scale pandemics such as COVID-19. In this study, we aimed to address the limitations of TCT. The Yale School of Engineering developed a Hardware-Assisted Bluetooth-based Infection Tracking (HABIT) device. Following the successful implementation of HABIT in a university setting, this study sought to evaluate the performance and implementation of HABIT in a high school setting using an embedded mixed methods design. In this pilot implementation study, we first assessed the performance of HABIT using mock case simulations in which we compared contact tracing data collected from mock case interviews (TCT) versus Bluetooth devices (HABIT). For each method, we compared the number of close contacts identified and identification of unique contacts. We then conducted an embedded mixed methods evaluation of the implementation outcomes of HABIT devices using pre- and postimplementation quantitative surveys and qualitative focus group discussions with users and implementers according to the Reach, Effectiveness, Adoption, Implementation, and Maintenance framework. In total, 17 students and staff completed mock case simulations in which 161 close contact interactions were detected by interview or Bluetooth devices. We detected significant differences in the number of close contacts detected by interview versus Bluetooth devices (P<.001), with most (127/161, 78.9%) contacts being reported by interview only. However, a significant number (26/161, 16.1%; P<.001) of contacts were uniquely identified by Bluetooth devices. The interface, ease of use, coherence, and appropriateness were highly rated by both faculty and students. HABIT provided emotional security to users. However, the prototype design and technical difficulties presented barriers to the uptake and sustained use of HABIT. Implementation of HABIT in a high school was impeded by technical difficulties leading to decreased engagement and adherence. Nonetheless, HABIT identified a significant number of unique contacts not reported by interview, indicating that electronic technologies may augment traditional contact tracing once user preferences are accommodated and technical glitches are overcome. Participants indicated a high degree of acceptance, citing emotional reassurance and a sense of security with the device.

Many have proposed the use of Bluetooth technology to help scale up contact tracing for COVID-19. However, much remains unknown about the accuracy of this technology in real-world settings, the attitudes of potential users, and the differences between delivery formats (mobile app vs carriable or wearable devices). We pilot tested 2 separate Bluetooth contact tracing technologies on a university campus to evaluate their sensitivity and specificity, and to learn from the experiences of the participants. We used a convergent mixed methods study design, and participants included graduate students and researchers working on a university campus during June and July 2020. We conducted separate 2-week pilot studies for each Bluetooth technology. The first was for a mobile phone app (\"app pilot\"), and the second was for a small electronic \"tag\" (\"tag pilot\"). Participants validated a list of Bluetooth-identified contacts daily and reported additional close contacts not identified by Bluetooth. We used these data to estimate sensitivity and specificity. Participants completed a postparticipation survey regarding appropriateness, usability, acceptability, and adherence, and provided additional feedback via free text. We used tests of proportions to evaluate differences in survey responses between participants from each pilot, paired t tests to measure differences between compatible survey questions, and qualitative analysis to evaluate the survey's free-text responses. Among 25 participants in the app pilot, 53 contact interactions were identified by Bluetooth and an additional 61 by self-report. Among 17 participants in the tag pilot, 171 contact interactions were identified by Bluetooth and an additional 4 by self-report. The tag had significantly higher sensitivity compared with the app (46/49, 94% vs 35/61, 57%; P<.001), as well as higher specificity (120/126, 95% vs 123/141, 87%; P=.02). Most participants felt that Bluetooth contact tracing was appropriate on campus (26/32, 81%), while significantly fewer participants felt that using other technologies, such as GPS or Wi-Fi, was appropriate (17/31, 55%; P=.02). Most participants preferred technology developed and managed by the university rather than a third party (27/32, 84%) and preferred not to have tracing apps on their personal phones (21/32, 66%), due to \"concerns with privacy.\" There were no significant differences in self-reported adherence rates across pilots. Convenient and carriable Bluetooth technology may improve tracing efficiency while alleviating privacy concerns by shifting data collection away from personal devices. With accuracy comparable to, and in this case, superior to, mobile phone apps, such approaches may be suitable for workplace or school settings with the ability to purchase and maintain physical devices.

We present a biomimetic system that captures essential functional properties of the glomerular layer of the mammalian olfactory bulb, specifically including its capacity to decorrelate similar odor representations without foreknowledge of the statistical distributions of analyte features. Our system is based on a digital neuromorphic chip consisting of 256 leaky-integrate-and-fire neurons, 1024 × 256 crossbar synapses, and address-event representation communication circuits. The neural circuits configured in the chip reflect established connections among mitral cells, periglomerular cells, external tufted cells, and superficial short-axon cells within the olfactory bulb, and accept input from convergent sets of sensors configured as olfactory sensory neurons. This configuration generates functional transformations comparable to those observed in the glomerular layer of the mammalian olfactory bulb. Our circuits, consuming only 45 pJ of active power per spike with a power supply of 0.85 V, can be used as the first stage of processing in low-power artificial chemical sensing devices inspired by natural olfactory systems.

Inspired by the brain's structure, we have developed an efficient, scalable, and flexible non–von Neumann architecture that leverages contemporary silicon technology. To demonstrate, we built a 5.4-billion-transistor chip with 4096 neurosynaptic cores interconnected via an intrachip network that integrates 1 million programmable spiking neurons and 256 million configurable synapses. Chips can be tiled in two dimensions via an interchip communication interface, seamlessly scaling the architecture to a cortexlike sheet of arbitrary size. The architecture is well suited to many applications that use complex neural networks in real time, for example, multiobject detection and classification. With 400-pixel-by-240-pixel video input at 30 frames per second, the chip consumes 63 milliwatts.

Emergent nanoscale non-volatile memory technologies with high integration density offer a promising solution to overcome the scalability limitations of CMOS-based neural networks architectures, by efficiently exhibiting the key principle of neural computation. Despite the potential improvements in computational costs, designing high-performance on-chip communication networks that support flexible, large-fanout connectivity remains as daunting task. In this paper, we elaborate on the communication requirements of large-scale neuromorphic designs, and point out the differences with the conventional network-on-chip architectures. We present existing approaches for on-chip neuromorphic routing networks, and discuss how new memory and integration technologies may help to alleviate the communication issues in constructing next-generation intelligent computing machines.

On-chip analog Ising Machines (IMs) are a promising means to solve difficult combinatorial optimization problems. For scalable on-chip realizations to be practical, 1) the problem should map scalably to Ising form, 2) interconnectivity between spins should be sparse, 3) the number of bits of coupling resolution (BCR) needed for programming interconnection weights should be small, and 4) the chip should be capable of solving problems with different connection topologies. We explore these issues for the SATisfiability problem and devise FPIM, a reconfigurable on-chip analog Ising machine scheme well suited for SAT. To map SAT problems onto FPIMs, we leverage Boolean logic synthesis as a first step, but replace synthesized logic gates with Ising equivalent circuits whose analog dynamics solve SAT by minimizing the Ising Hamiltonian. We apply our approach to 2000 benchmark problems from SATLIB,demonstrating excellent scaling, together with low sparsity and low BCR that are independent of problem scale. Placement/routing reveals a very feasible requirement of less than 10 routing tracks to implement all the benchmarks, translating to an area requirement of about 10mm^2 for a programmable 1000-spin FPIM in 65nm technology.

Mutual exclusion is an important problem in the context of shared resource usage, where only one process can be using the shared resource at any given time. A mutual exclusion protocol that does not use information on the duration for which each process uses the resource can lead to sub-optimal utilization times. We consider a simple two-process mutual exclusion problem with a central server that provides access to the shared resource. We show that even in the absence of a clock, under certain conditions, the server can opportunistically grant early access to a client based on timing information. We call our new protocol opportunistic mutual exclusion. Our approach requires an extra request signal on each channel between client and server to convey extra information, and the server can grant early access based only on the order of events rather than through measuring time. We derive the handshaking specification and production rules for our protocol, and report on the energy and delay of the circuits in a 65nm process.

Inspired by the brain's structure, we have developed an efficient, scalable, and flexible non-von Neumann architecture that leverages contemporary silicon technology. To demonstrate, we built a 5.4-billion-transistor chip with 4096 neurosynaptic cores interconnected via an intrachip network that integrates 1 million programmable spiking neurons and 256 million configurable synapses. Chips can be tiled in two dimensions via an interchip communication interface, seamlessly scaling the architecture to a cortexlike sheet of arbitrary size. The architecture is well suited to many applications that use complex neural networks in real time, for example, multiobject detection and classification. With 400-pixel-by-240-pixel video input at 30 frames per second, the chip consumes 63 milliwatts.