Abstract: Multistate non-volatile photonic memory devices are disclosed. The photonic devices comprise phase change materials with broadband transparencies used to store discretized information with negligible losses in the 0 state. The photonic memories comprise multiple configurations for reading and writing multi-bit words. The reading mechanisms comprises schemes based on light-absorption (FIG. 1), shift in resonances of a cavity (ring resonator, photonic crystal; FIG. 2) or interferometric schemes (FIG. 3). The photonic memory devices employ multiple techniques for writing electrically (FIG. 4 and related performance) and/or all-optically (FIGS. 7-10). The optical writing can be performed with pulsed laser light coming either from free space or on-chip using dedicated writing lines and opportune drops.

Abstract: A display may overlap a sensor such as a camera or ambient light sensor. A portion of the display that overlaps the sensor may be modified to increase transparency relative to the remaining portion of the display. The modified portion of the display may have pixel islands that conform to a regular grid pattern of the remaining portion of the display. One or more gate lines and/or one or more data lines may pass through the modified portion of the display without connecting to any subpixels in the modified portion of the display. An opaque pixel definition layer may include some openings that are aligned with opaque masking layer openings to allow on-axis light to pass through to a sensor and some openings that are offset relative to respective opaque masking layer openings to allow off-axis light to pass through to the sensor.

Abstract: A light emitter that operates through a display may cause display artifacts, even when the light emitter operates using non-visible wavelengths. Display artifacts caused by a light emitter that operates through a display may be referred to as emitter artifacts. To mitigate emitter artifacts, a display pixel that overlaps the light emitter may include a transistor that provides a current sink to divert leakage current away from the light-emitting diode in that pixel. A shielding layer may be interposed between the light emitter and a display pixel. The shielding layer may block light from the light emitter. The shielding layer may have an opening that exposes one transistor in the pixel to the light from the light emitter. Pixels that overlap the light emitter may have larger anodes than pixels that do not overlap the light emitter.

Abstract: Systems and methods described herein may utilize a non-linear scaling relationship to scale brightness of selective portions of the display in a manner that reduces or eliminates perceivable banding effects. By non-linearly controlling changes in brightness of the display, a viewer may perceive a more uniform, linear dimming towards the relatively dimmer region without perceiving banding, leading to improved user experience when viewing the electronic display.

Abstract: An electronic device may include a display panel implementing a first region having a first pixel layout and a second region having a second pixel layout, a sensor disposed behind the first region, and image processing circuitry communicatively coupled to the display panel. The image processing circuitry may process image data associated with the first region of the display panel using a first pixel layout resampler and process image data associated with the second region of the display panel using a second pixel layout resampler.

Abstract: In a display characterized by regions with different pixel responses due, for example, to local pixel density variation, voltage-to-luminance matching may be non-universal. Therefore, in order to avoid visual artifacts that may hinder a desired visualization of displayed content, it may be advantageous to compensate the different gamma responses. In some cases, such as with electronic devices having a single pixel density across the display, optical calibration may be performed to determine voltage-to-luminance matching. However, in electronic devices with local pixel density variations, it may be disadvantageous to perform optical calibrations for each region with a different pixel density. Instead of using two distinct gamma curves which may include dedicated optical calibration, a global nonlinear scaler (GNLS) compensation may be applied.

Abstract: Systems and methods described herein may utilize a non-linear scaling relationship to scale brightness of selective portions of the display in a manner that reduces or eliminates perceivable banding effects. By non-linearly controlling changes in brightness of the display, a viewer may perceive a more uniform, linear dimming towards the relatively dimmer region without perceiving banding, leading to improved user experience when viewing the electronic display.

Abstract: In a display characterized by regions with different pixel responses due, for example, to local pixel density variation, voltage-to-luminance matching may be non-universal. Therefore, in order to avoid visual artifacts that may hinder a desired visualization of displayed content, it may be advantageous to compensate the different gamma responses. In some cases, such as with electronic devices having a single pixel density across the display, optical calibration may be performed to determine voltage-to-luminance matching. However, in electronic devices with local pixel density variations, it may be disadvantageous to perform optical calibrations for each region with a different pixel density. Instead of using two distinct gamma curves which may include dedicated optical calibration, a global nonlinear scaler (GNLS) compensation may be applied.

Abstract: Amplitude-only Fourier optical processors is capable of processing large-scale matrices in a single time-step and microsecond-short latency. The processors may have a 4f optical system architecture and may employ reprogrammable high-resolution amplitude-only spatial modulators, such as Digital Micromirror Devices (DMD). In addition, methods are provided for obtaining amplitude-only electro-optical convolutions between large matrices displayed by the DMDs. The large matrices on which convolution is performed may be feature maps corresponding to images and kernel matrices used in neural networks classification systems. Analog optical convolutional neural networks are also provided that perform accurate classification tasks on large matrices. In addition, methods are provided for off-chip training the analog optical convolutional neural networks.

Abstract: An accelerator for modern convolutional neural networks applies the Winograd filtering algorithm in a wavelength division multiplexing integrated photonics circuit modulated by a memristor-based analog memory unit.

Abstract: A system performing optical and/or electro-optical tensor operations and featuring a photonic dot product engine with a first input and a second input and summation to perform multiply-accumulate operations. The first and/or second input is a matrix, and/or a vector, and/or scalar. The system is a Photonic Tensor Core.

Abstract: An accelerator for modern convolutional neural networks applies the Winograd filtering algorithm in a wavelength division multiplexing integrated photonics circuit modulated by a memristor-based analog memory unit.

Abstract: An accelerator for modern convolutional neural networks applies the Winograd filtering algorithm in a wavelength division multiplexing integrated photonics circuit modulated by a memristor-based analog memory unit.

Abstract: An accelerator for modern convolutional neural networks applies the Winograd filtering algorithm in a wavelength division multiplexing integrated photonics circuit modulated by a memristor-based analog memory unit.