Convolutional Neural Networks (CNNs) are widely adopted in object recognition, speech processing and machine translation, due to their extremely high inference accuracy. However, it is challenging to compute massive computationally expensive convolutions of deep CNNs on traditional CPUs and GPUs. Emerging Nanophotonic technology has been employed for on-chip data communication, because of its CMOS compatibility, high bandwidth and low power consumption. In this paper, we propose a nanophotonic accelerator, HolyLight, to boost the CNN inference throughput in datacenters. Instead of an all-photonic design, HolyLight performs convolutions by photonic integrated circuits, and process the other operations in CNNs by CMOS circuits for high inference accuracy. We first build HolyLight-M by microdisk-based matrix-vector multipliers. We find analog-todigital converters (ADCs) seriously limit its inference throughput per Watt. We further use microdisk-based adders and shifters to architect HolyLight-A without ADCs. Compared to the stateof-the-art ReRAM-based accelerator, HolyLight-A improves the CNN inference throughput per Watt by 13× with trivial accuracy degradation.
Crosstalk noise is an intrinsic characteristic of photonic devices used by optical networks-on-chip (ONoCs) as well as a potential issue. For the first time, this paper analyzed and modeled the crosstalk noise, signal-to-noise ratio (SNR), and bit error rate (BER) of optical routers and ONoCs. The analytical models for crosstalk noise, minimum SNR, and maximum BER in meshbased ONoCs are presented. An automated crosstalk analyzer for optical routers is developed. We find that crosstalk noise significantly limits the scalability of ONoCs. For example, due to crosstalk noise, the maximum BER is 10 -3 on the 8×8 meshbased ONoC using an optimized crossbar-based optical router. To achieve the BER of 10 -9 for reliable transmissions, the maximum ONoC size is 6×6. A novel compact high-SNR optical router is proposed to improve the maximum ONoC size to 8×8.
BackgroundWithin the domain of craniomaxillofacial surgery, orthognathic surgery is a special field dedicated to the correction of dentofacial anomalies resulting from skeletal malocclusion. Generally, in such cases, an interdisciplinary orthodontic and surgical treatment approach is required. After initial orthodontic alignment of the dental arches, skeletal discrepancies of the jaws can be corrected by distinct surgical strategies and procedures in order to achieve correct occlusal relations, as well as facial balance and harmony within individualized treatment concepts. To transfer the preoperative surgical planning and reposition the mobilized dental arches with optimal occlusal relations, surgical splints are typically used. For this purpose, different strategies have been described which use one or more splints. Traditionally, these splints are manufactured by a dental technician based on patient-specific dental casts; however, computer-assisted technologies have gained increasing importance with respect to preoperative planning and its subsequent surgical transfer.Methods: In a pilot study of 10 patients undergoing orthognathic corrections by a one-splint strategy, two final occlusal splints were produced for each patient and compared with respect to their clinical usability. One splint was manufactured in the traditional way by a dental technician according to the preoperative surgical planning. After performing a CBCT scan of the patient’s dental casts, a second splint was designed virtually by an engineer and surgeon working together, according to the desired final occlusion. For this purpose, RapidSplint®, a custom-made software platform, was used. After post-processing and conversion of the datasets into .stl files, the splints were fabricated by the PolyJet procedure using photo polymerization. During surgery, both splints were inserted after mobilization of the dental arches then compared with respect to their clinical usability according to the occlusal fitting.ResultsUsing the workflow described above, virtual splints could be designed and manufactured for all patients in this pilot study. Eight of 10 virtual splints could be used clinically to achieve and maintain final occlusion after orthognathic surgery. In two cases virtual splints were not usable due to insufficient occlusal fitting, and even two of the traditional splints were not clinically usable. In five patients where both types of splints were available, their occlusal fitting was assessed as being equivalent, and in one case the virtual splint showed even better occlusal fitting than the traditional splint. In one case where no traditional splint was available, the virtual splint proved to be helpful in achieving the final occlusion.ConclusionsIn this pilot study it was demonstrated that clinically usable splints for orthognathic surgery can be produced by computer-assisted technology. Virtual splint design was realized by RapidSplint®, an in-house software platform which might contribute in future to shorten preoperative workflows fo...
Class III myosins (Myo3) and actin-bundling protein Espin play critical roles in regulating the development and maintenance of stereocilia in vertebrate hair cells, and their defects cause hereditary hearing impairments. Myo3 interacts with Espin1 through its tail homology I motif (THDI), however it is not clear how Myo3 specifically acts through Espin1 to regulate the actin bundle assembly and stabilization. Here we discover that Myo3 THDI contains a pair of repeat sequences capable of independently and strongly binding to the ankyrin repeats of Espin1, revealing an unexpected Myo3-mediated cross-linking mechanism of Espin1. The structures of Myo3 in complex with Espin1 not only elucidate the mechanism of the binding, but also reveal a Myo3-induced release of Espin1 auto-inhibition mechanism. We also provide evidence that Myo3-mediated cross-linking can further promote actin fiber bundling activity of Espin1.DOI:
http://dx.doi.org/10.7554/eLife.12856.001
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.