LOW-POWER EDGE COMPUTING WITH OPTICAL NEURAL NETWORKS VIA WDM WEIGHT BROADCASTING

    公开(公告)号:WO2022086615A2

    公开(公告)日:2022-04-28

    申请号:PCT/US2021/043593

    申请日:2021-07-29

    Abstract: NetCast is an optical neural network architecture that circumvents constraints on deep neural network (DNN) inference at the edge. Many DNNs have weight matrices that are too large to run on edge processors, leading to limitations on DNN inference at the edge or bandwidth bottlenecks between the edge and server that hosts the DNN. With NetCast, a weight server stores the DNN weight matrix in local memory, modulates the weights onto different spectral channels of an optical carrier, and distributes the weights to one or more clients via optical links. Each client stores the activations, or layer inputs, for the DNN and computes the matrix-vector product of those activations with the weights from the weight server in the optical domain. This multiplication can be performed coherently by interfering the spectrally multiplexed weights with spectrally multiplexed activations or incoherently by modulating the weight signal from the weight server with the activations.

    MULTIPLEXED SENSOR NETWORK USING SWEPT SOURCE RAMAN SPECTROSCOPY

    公开(公告)号:WO2022026053A1

    公开(公告)日:2022-02-03

    申请号:PCT/US2021/035105

    申请日:2021-06-01

    Abstract: Swept-source Raman spectroscopy uses a tunable laser and a fixed-wavelength detector instead of a spectrometer or interferometer to perform Raman spectroscopy with the throughput advantage of Fourier transform Raman spectroscopy without bulky optics or moving mirrors. Although the tunable laser can be larger and more costly than a fixed wavelength diode laser used in other Raman systems, it is possible to split and switch the laser light to multiple ports simultaneously and/or sequentially. Each site can be monitored by its own fixed-wavelength detector. This architecture can be scaled by cascading fiber switches and/or couplers between the tunable laser and measurement sites. By multiplexing measurements at different sites, it is possible to monitor many sites at once. Moreover, each site can be meters to kilometers from the tunable laser. This makes it possible to perform swept-source Raman spectroscopy at many points across a continuous flow manufacturing environment with a single laser.

    PASSIVE WIDE-AREA THREE-DIMENSIONAL IMAGING
    93.
    发明申请

    公开(公告)号:WO2021101612A2

    公开(公告)日:2021-05-27

    申请号:PCT/US2020/048334

    申请日:2020-08-28

    Inventor: SWEENEY, Robert

    Abstract: Radar, lidar, and other active 3D imaging techniques require large, heavy sensors that consume lots of power. Passive 3D imaging techniques based on feature matching are computationally expensive and limited by the quality of the feature matching. Fortunately, there is a robust, computationally inexpensive way to generate 3D images from full-motion video acquired from a platform that moves relative to the scene. The full-motion video frames are registered to each other and mapped to the scene coordinates using data about the trajectory of the platform with respect to the scene. The time derivative of the registered frames equals the product of the height map of the scene, the projected angular velocity of the platform, and the spatial gradient of the registered frames. This relationship can be solved in (near) real time to produce the height map of the scene from the full-motion video and the trajectory.

    FREELY SCALABLE QUANTUM COMPUTING USING A 2D ATOMIC EMITTER ARRAY WITH MASSIVELY PARALLEL OPTICAL INTERCONNECTS

    公开(公告)号:WO2021076221A1

    公开(公告)日:2021-04-22

    申请号:PCT/US2020/046729

    申请日:2020-08-18

    Abstract: Quantum information processing involves entangling large numbers of qubits, which can be realized as defect centers in a solid-state host. The qubits can be implemented as individual unit cells, each with its own control electronics, that are arrayed in a cryostat. Free- space control and pump beams address the qubit unit cells through a cryostat window. The qubit unit cells emit light in response to these control and pump beams and microwave pulses applied by the control electronics. The emitted light propagates through free space to a mode mixer, which interferes the optical modes from adjacent qubit unit cells for heralded Bell measurements. The qubit unit cells are small (e.g., 10 pm square), so they can be tiled in arrays of up to millions, addressed by free- space optics with mi cron- scale spot sizes. The processing overhead for this architecture remains relatively constant, even with large numbers of qubits, enabling scalable large-scale quantum information processing.

    NON-INVASIVE GLUCOSE MONITORING BY RAMAN SPECTROSCOPY

    公开(公告)号:WO2021040878A1

    公开(公告)日:2021-03-04

    申请号:PCT/US2020/040078

    申请日:2020-06-29

    Abstract: Noninvasive glucose monitoring has been a long-standing need in diabetes management. Among many approaches to meeting this need, Raman spectroscopy has attracted attention due to its molecular specificity. Previous Raman-based glucose sensing can predict blood glucose concentration based on a statistical correlation between the reference glucose concentration and unspecified spectral features. However, the lack of glucose Raman peaks and non-prospective prediction have led to questions about the effectiveness of in vivo Raman spectroscopy for transcutaneous glucose sensing. Here, we disclose technology for directly observing distinct glucose Raman spectra from skin. The Raman signal intensities were proportional to the reference glucose concentrations in three live swine glucose clamping experiments. Tracking the spectral intensity based on the linearity enables prospective prediction with high accuracy in within-subject and inter-subject models. Compared to previous statistical approaches, prospective predictions based on a direct glucose signal from the skin offers robust, reliable noninvasive glucose sensing.

    IDENTIFYING AND RETRIEVING VIDEO METADATA WITH PERCEPTUAL FRAME HASHING

    公开(公告)号:WO2021003323A1

    公开(公告)日:2021-01-07

    申请号:PCT/US2020/040584

    申请日:2020-07-02

    Abstract: Shoppable video enables a viewer to identify and buy items appearing in a video. To retrieve information about the items in a frame of the video, the playback device generates a perceptual hash of that frame and uses that hash to query a first database storing perceptual hashes of different version of the video. The database query returns an identifier for the frame, which is then used to query a second database that store the item information. The results of this query are returned to the playback device, which shows them to the user, enabling the viewer to learn more about and possibly purchase the item. Using queries based on perceptual hashes of different versions of the video increases the likelihood of returning a match, despite formatting differences. And using separate hash and metadata databases makes it possible to update the metadata without changing the hashes.

    METHODS AND APPARATUS FOR LOW-VOLATILITY SAMPLING

    公开(公告)号:WO2020068280A1

    公开(公告)日:2020-04-02

    申请号:PCT/US2019/045661

    申请日:2019-08-08

    Abstract: Molecular rotational resonance (MRR) spectroscopy is a structurally-specific, high-resolution spectroscopy technique that can provide accurate reaction process data with finer time resolution than existing techniques. It is the only analytical technique that can make online chiral composition measurements. This makes it especially useful for online reaction monitoring, which is done today by manually pulling off samples and measuring samples offline and takes 3-4 hours per measurement. Conversely, an MRR spectrometer can resolve isomers in about 10 minutes when fed with a low-volatility sampling interface that connects directly to the reaction line. The sampling interface measures a precise sample of the reaction solution, boils off the solvent to concentrate the analyte, volatilizes the analyte, and injects the volatilized analyte into the MRR spectrometer's measurement chamber for an MRR measurement. The sample concentration and volatilization happen quickly and without any extra sample preparation.

    TUNABLE GRAPHENE METAMATERIALS FOR BEAM STEERING AND TUNABLE FLAT LENSES

    公开(公告)号:WO2020036626A2

    公开(公告)日:2020-02-20

    申请号:PCT/US2019/016751

    申请日:2019-02-06

    Abstract: Modulating graphene's optical conductivity with an electrolyte nanopatterning technique reduces or eliminates scattering loss caused by rough edges from etching. This technique uses a resist mask patterned with features as small as 30 nm to shield graphene from ions in an electrolyte. It can provide a carrier density variation of about 10 14 cm -2 across a length of just 15 nm. And it can be combined with a technique of growing or transferring graphene on atomically smooth hexagonal boron nitride (hBN) to increase graphene's carrier mobility, e.g., to 10,000 cm 2 /(V.s) or more. The resulting graphene metamaterials can be used to make voltage-tunable electro- optical devices, such as beam-steering devices, electro-optical switch and modulators, and reconfigurable holograms.

Patent Agency Ranking