Abstract: An optimized computer architecture for training an neural network includes a system having multiple GPUs. The neural network may be divided into separate portions, and a different portion is assigned to each of the multiple GPUs. Within each GPU, its portion is further divided across multiple training worker threads in multiple processing cores, and each processing core has lock-free access to a local parameter memory. The local parameter memory of each GPU is separately, and individually, synchronized with a remote master parameter memory by lock memory access. Each GPU has a separate set of communication worker threads dedicated to data transfer between the GPU and the remote parameter memory so that the GPU's training worker threads are not involved with cross GPU communications.

Abstract: Systems, methods, and non-transitory computer-readable media can receive a plurality of web templates associated with a web page. A buffer having a fixed, pre-determined length is allocated. A first set of data associated with a first web template of the plurality of web templates is copied to the buffer. A second set of data associated with a second web template of the plurality of web templates is copied to the buffer.

Abstract: A method includes providing light from a light source and separating the light into a first portion of the light and a second portion of the light that is spatially separated from the first portion of the light. The method also includes transmitting the first portion of the light through a first set of optical elements to provide a first wide-field beam, transmitting the second portion of the light through a second set of optical elements to provide a second wide-field beam that is spatially separated from the first wide-field beam, and transmitting the second wide-field beam through a third set of optical elements to provide a plurality of separate light patterns. The method further includes concurrently projecting the first wide-field beam and the plurality of separate light patterns onto an optically recordable medium to form a holographic medium.

Abstract: An online system maintains one or more models that determine likelihoods of a user performing various interactions after being presented with a content item. Additionally, the online system receives information identifying interactions by users with content, and generates embeddings for various users based on the interactions by the users with content. When determining whether to present a content item including an objective identifying an interaction to a user, the online system applies a maintained model to determine a likelihood of the user performing the interaction identified by the objective after being presented with the content item. Additionally, the online system determines a similarity of the embedding of the user to embeddings of users who performed the interaction identified by the objective. Based on a combination of the likelihood determined by the model and the similarity, the online system determines whether to present the content item to the user.

Abstract: A haptic-jamming device may include (1) a first haptic-jamming member having a polarizable first element rotatable about an axis and a first elongated extension outwardly disposed from the first element in a first direction, (2) a second haptic-jamming member having a second element rotatable about the axis and a second elongated extension outwardly disposed from the second element in a second direction, and (3) a variable voltage source for at least one of the first haptic-jamming member or the second haptic-jamming member that, when energized, generates an electrostatic field encompassing at least one of the first or second elements. A level of separation of the first and second elements may be related to a strength of the electrostatic field. Various other methods, systems, and wearable articles are also disclosed.

Abstract: Systems and methods for creating digital interactive metadata elements in a social media composition may include (1) providing, to a user of a social media application, a composition interface for creating digital social media compositions, (2) collecting, via the composition interface, user-generated content and metadata information, (3) creating, based on the user-generated content and the metadata information, a social media composition that includes the user-generated content and an interactive metadata item, and (4) posting the social media composition to a consumption channel (e.g., a newsfeed channel and/or a stories channel) of a social-networking platform associated with the social media application. Various other methods, systems, and computer-readable media are also described.

Abstract: A haptic communication device includes a speech signal generator configured to receive speech sounds or a textual message and generate speech signals corresponding to the speech sounds or the textual message. An envelope encoder is operably coupled to the speech signal generator to extract a temporal envelope from the speech signals. The temporal envelope represents changes in amplitude of the speech signals. Carrier signals having a periodic waveform are generated. Actuator signals are generated by encoding the changes in the amplitude of the speech signals from the temporal envelope into the carrier signals. One or more cutaneous actuators are operably coupled to the envelope encoder to generate haptic vibrations representing the speech sounds or the textual message using the actuator signals.

Abstract: A layered optical element includes a substrate layer, an electrode layer disposed on the substrate layer, a liquid crystal (LC) layer comprising LC molecules, and a nanopatterned alignment layer in physical contact with the LC layer and disposed on a surface of either the substrate layer or the electrode layer. The nanopatterned alignment layer includes an arrangement of nanostructures, e.g., a grouping of nanolines. For a subset of the grouping of nanolines, the nanolines are configured to orient the LC molecules along a varying local orientation direction of each of nanoline in the subset. The varying local orientation direction of each nanoline in the subset can vary along a length of each nanoline.

Abstract: An optical device (e.g., a pupil expander) includes a waveguide with a slanted facet. The optical device includes a reflector on the slanted facet and a prism, or a grating at the slanted facet. The prism or the grating compensates for the dispersion of an image light from a display, which reduces smearing of displayed images. The waveguide can be configured for pupil replication in one-dimension or two-dimensions.

Abstract: A pupil-replicating waveguide suitable for operation with a coherent light source is disclosed. A waveguide body has opposed surfaces for guiding a beam of image light. An out-coupling element is disposed in an optical path of the beam for out-coupling portions of the beam at a plurality of spaced apart locations along the optical path. Electrodes are coupled to at least a portion of the waveguide body for modulating an optical path length of the optical path of the beam to create time-varying phase delays between the portions of the beam out-coupled by the out-coupling element.

Abstract: In one embodiment, a method includes accessing a notification to be sent to a user, where the notification has a context. The method also includes sending a request to a history service for historical notification data associated with the user with respect to the context of the notification and a ranking of the notification where the ranking indicates a probability of the user interacting with the notification. The method also includes receiving the historical notification data and the ranking from the history service. Moreover, the method also includes determining a delivery policy to apply to the notification based at least in part on the context of the notification, the historical notification data, and the ranking. Furthermore, the method also includes applying the delivery policy to the notification to be sent to the user.

Abstract: A method includes adjusting a focal length of an adaptive liquid lens that includes a layer of optical fluid on a substrate and in conjunction with adjusting the focal length of the adaptive liquid lens, adjusting a liquid crystal element coupled with the adaptive liquid lens. The method also includes transmitting light through the adaptive liquid lens and the liquid crystal element. Also disclosed is a hybrid lens including an adaptive liquid lens and a liquid crystal element coupled with the adaptive liquid lens. The liquid crystal element reduces optical artifacts caused by the adaptive liquid lens. The adaptive liquid lens includes a layer of optical fluid on a substrate, where a focal length of the adaptive liquid lens is adjustable.

Abstract: An illumination assembly includes a transparent substrate and a plurality of micro devices. The transparent substrate includes a first surface and a second surface opposite the first surface. The first surface includes a viewing region through which light passes prior to reaching an eyebox. The plurality of micro devices are coupled to respective conductive pathways that are affixed to the first surface. The plurality of micro devices including at least one micro device that is positioned within the viewing region. In some embodiments, the conductive pathways are arranged in a pseudo random manner. In some instances, the viewing region is composed of a circuity free region that is circumscribed by an outer region, and the plurality of micro devices are coupled to respective conductive pathways that are affixed to the first surface in the outer region.

Abstract: In one embodiment, a method includes accessing a first video; predicting a first noteworthy portion for the first video, wherein the first noteworthy portion is a portion of the first video that is predicted based on user-engagement information associated with the portion of the first video; extracting a first highlight from the first video, wherein the first highlight corresponds to the first noteworthy portion; and sending, to a client system of a user, information configured to render the first highlight and a first interactive element that is configured to launch the first video on the client system.

Abstract: The disclosed projector device may include (1) a first monochromatic emitter array having a plurality of emitters of a first color disposed in a two-dimensional configuration and (2) a second monochromatic emitter array having a plurality of emitters of a second color disposed in a two-dimensional configuration. The first and second monochromatic emitter arrays may be configured to emit images of the first and second colors into a waveguide configuration, and the first color may be different than the second color. Associated display systems and methods are also provided.

Abstract: A content review system for an online system automatically determines if received content items to be displayed to users violate any policies of the online system. The content review system generates a semantic vector representing the semantic features of a content item, for example, using a neural network. By comparing the semantic vector for the content item with semantic vectors of content items previously determined to violate one or more policies, the content review system determines whether the content item also violates one or more policies. The content review system may also maintain templates corresponding to portions of semantic vectors shared by multiple content items. An analysis of historical content items that conform to the template is performed to determine a probability that received content items that conform to the template violate a policy.

Abstract: A controller for a tiltable MEMS reflector is configured to oscillate the reflector about X axis or about X and Y axes, and to obtain information about current and past tilt angles. The controller is configured to evaluate tilt angles of the tiltable MEMS reflector at a later moment of time based on the previously obtained information about the tilt angles of the tiltable MEMS reflector at the different earlier moments of time. The controller may be further configured to energize the light source providing a light beam to the tiltable MEMS reflector at the later moment of time with brightness and color corresponding to the brightness and color of a pixel that will be painted by the tiltable MEMS reflector at the later moment of time. A statistical model may be combined with machine learning to accurately predict future tilt angles of the tiltable MEMS reflector.