Abstract: According to one embodiment, a semiconductor device includes a first conductive part, a semiconductor element, a first terminal, and a metal layer. The semiconductor element is located on the first conductive part. The first terminal is separated from the first conductive part in a second direction perpendicular to a first direction. The first direction is from the first conductive part toward the semiconductor element. The first terminal includes a first portion, and a second portion located between the first portion and the first conductive part. A lower surface of the second portion is positioned lower than a lower surface of the first portion and lower than a lower surface of a first insulating portion. The first insulating portion is located between the first conductive part and the second portion. The metal layer is located at the lower surfaces of the first portion and of the second portion.

Abstract: A brain-inspired cognitive learning method can obtain good learning results in various environments and tasks by selecting the most suitable algorithm models and parameters based on the environments and tasks, and can correct wrong behavior. The framework includes four main modules: a cognitive feature extraction module, a cognitive control module, a learning network module, and a memory module. The memory module includes a data base, a cognitive case base, and an algorithm and hyper-parameter base, which store data of dynamic environments and tasks, cognitive cases, and concrete algorithms and hyper-parameter values, respectively. For dynamic environments and tasks, the most suitable algorithm model and hyper-parameter combination can be flexibly selected. In addition, with “good money drives out bad”, mislabeled data is corrected using correctly labeled data, to achieve robustness of training data.

Abstract: Some embodiments provide a compiler for optimizing the implementation of a machine-trained network (e.g., a neural network) on an integrated circuit (IC). The compiler of some embodiments receives a specification of a machine-trained network including multiple layers of computation nodes and generates a graph representing options for implementing the machine-trained network in the IC. The compiler, as part of generating the graph, in some embodiments, determines whether any set of channels contains no non-zero values (i.e., contains only zero values). For sets of channels that include no non-zero values, some embodiments perform a zero channel removal operation to remove all-zero channels wherever possible. In some embodiments, zero channel removal operations include removing input channels, removing output channels, forward propagation, and backward propagation of channels and constants.

Abstract: The present disclosure provides an effect application previewing method, an apparatus, a device and a storage medium. The method includes: taking, in response to a preview trigger operation for a target effect style, a position of a pointer on a video track as a start point, to generate a virtual video frame for the target effect style; synchronously playing the virtual video frame and at least one video clip corresponding to the virtual video frame on the video track based on a timeline, to preview an effect of the target effect style applied to the at least one video clip.

Abstract: A feedforward generative neural network that generates an output example that includes multiple output samples of a particular type in a single neural network inference. Optionally, the generation may be conditioned on a context input. For example, the feedforward generative neural network may generate a speech waveform that is a verbalization of an input text segment conditioned on linguistic features of the text segment.

Abstract: Systems and processes for operating a digital assistant are provided. An example method includes, at an electronic device having one or more processors and memory, receiving an audio input including an utterance, determining, based on a speaker profile, an identity of a speaker of the utterance, determining whether the identity of the speaker matches a predetermined identity, and in accordance with a determination that the identity of the speaker matches the predetermined identity selectively adjusting a volume of the utterance relative to a volume of other sound of the audio input and providing an output of the adjusted utterance.

Abstract: The present disclosure is related to a method to generate user representative avatars that fit within a design paradigm. The method includes receiving depth information corresponding to multiple user features of the user, determining one or more feature landmarks for the user based on the depth information, utilizing the one or more feature landmarks to classify a first user feature relative to an avatar feature category, selecting a first avatar feature from the avatar feature category based on the classification of the first user feature, combining the first avatar feature within an avatar representation to generate a user avatar, and output the user avatar for display.

Abstract: A generator network of a variational autoencoder can be trained to approximate a simulator and generate a first result. The simulator is associated with input data, based on which the simulator outputs output data. A training data set for the generator network can include the simulator's input data and output data. Based on the simulator's output data and the first result of the generator network, an inference network of the variational autoencoder can be trained to generate a second result. The second result of the trained inference network inverts the first result of the generator and approximates the simulator's input data. The trained inference network can function as an inverted simulator.

Abstract: A mount type device includes a camera configured to photograph a work site in a wider range than a visual field of a worker; a sensor configured to obtain a visual line of the worker; and a visual line movement detector configured to detect, as visual line movement information, movement of a direction of the visual line of the worker by using the sensor information. The mount type device transmits data on the photographed image and the visual line movement information to an operation instructing apparatus via a network. An operation instructing apparatus includes a display setter configured to extract a first image having a wider range than a visual field range of the worker from the image photographed by the camera, correct the first image using the visual line movement information, and cause a display of the corrected first image; and an instruction image is generated for the worker.

Abstract: Methods, systems, and apparatus, including instructions encoded on storage media, for performing reduction of gradient vectors for distributed training of a neural network. One of the methods includes receiving, at each of the plurality of devices, a respective batch; performing, by each device, a forward pass comprising, for each batch normalization layer: generating, by each of the devices, a respective output of the corresponding other layer for each training example in the batch, determining, by each of the devices, a per-replica mean and a per-replica variance; determining, for each sub-group, a distributed mean and a distributed variance from the per-replica means and the per-replica variances for the devices in the sub-group; and applying, by each device, batch normalization to the respective outputs of the corresponding other layer generated by the device using the distributed mean and the distributed variance for the sub-group to which the device belongs.

Abstract: Methods, systems, and apparatus, including medium-encoded computer program products, for providing editable keyframe-based animation data for applying to a character to animate motion of the character in three-dimensional space. Three-dimensional motion data is constructed from two-dimensional videos. The three-dimensional motion data represents movement of people in the two-dimensional videos and includes, for each person, a root of a three-dimensional skeleton of the person. The three-dimensional skeleton comprises multiple three-dimensional poses of the person during at least a portion of frames of a video from the two-dimensional videos. The three-dimensional motion data is converted into editable keyframe-based animation data in three-dimensional space and provided to animate motion.

Abstract: An electronic apparatus for performing machine learning a method of machine learning, and a non-transitory computer-readable recording medium are provided. The electronic apparatus includes an operation module configured to include a plurality of processing elements arranged in a predetermined pattern and share data between the plurality of processing elements which are adjacent to each other to perform an operation; and a processor configured to control the operation module to perform a convolution operation by applying a filter to input data, wherein the processor controls the operation module to perform the convolution operation by inputting each of a plurality of elements configuring a two-dimensional filter to the plurality of processing elements in a predetermined order and sequentially applying the plurality of elements to the input data.

Abstract: According to an embodiment, a method includes estimating, based at least in part on a first location of a bounding box for a foot in a first frame of a sequence of frames and a second location of the bounding box in a second frame of the sequence of frames, a smoothing transformation that characterizes a movement of the foot in the sequence of frames and smoothing, based at least in part on the smoothing transformation, a movement of a structure of the foot from a first location of the structure in the first frame to a second location of the structure in the second frame. The method also includes superimposing a model of a shoe onto the foot in the sequence of frames based at least in part on the smoothed movement.

Abstract: A three-dimensional face model generation method is provided. The method includes: obtaining an inputted three-dimensional face mesh of a target object; aligning the three-dimensional face mesh with a first three-dimensional face model of a standard object according to face keypoints; performing fitting on the three-dimensional face mesh and a local area of the first three-dimensional face model, to obtain a second three-dimensional face model after local fitting; and performing fitting on the three-dimensional face mesh and a global area of the second three-dimensional face model, to obtain a three-dimensional face model of the target object after global fitting.

Abstract: An insulating component includes an insulating substrate, a metal layer, a bond, and a lead terminal. The plate-like insulating substrate has a groove continuous from its upper to side surfaces. The metal layer includes a first metal layer on the upper surface of the insulating substrate and a second metal layer on an inner surface of the groove continuous with the first metal layer. The bond is on an upper surface of the metal layer. The lead terminal is on an upper surface of the first metal layer with the bond in between, and overlaps the grooves. The bond includes a first bond fixing the lead terminal to the first metal layer and a second bond on an upper surface of the second metal layer continuous with the first bond. The groove includes an inner wall having a ridge. The second bond is between the ridge and the lead terminal.

Abstract: A gate-controlled diode includes a substrate, a gate stacked on the substrate, a gate insulation layer, a first two-dimensional semiconductor layer, a second two-dimensional semiconductor layer, a source, and a drain disposed separately from the source. The gate is embedded in a surface of the substrate, and the gate insulation layer covers the surface of the substrate in which the gate is disposed. The first two-dimensional semiconductor layer is stacked on the gate insulation layer, a portion of the second two-dimensional semiconductor layer is stacked on the gate insulation layer, another portion is stacked on the first two-dimensional semiconductor layer. The another portion of the second two-dimensional semiconductor layer stacked on the first two-dimensional semiconductor layer forms a heterojunction. An orthographic projection of the heterojunction onto the substrate is in an orthographic projection of the gate onto the substrate.

Abstract: The present disclosure describes a semiconductor structure that includes a channel region, a source region adjacent to the channel region, a drain region, a drift region adjacent to the drain region, and a dual gate structure. The dual gate structure includes a first gate structure over portions of the channel region and portions of the drift region. The dual gate structure also includes a second gate structure over the drift region.

Abstract: A multiply-accumulate method and architecture are disclosed. The architecture includes a plurality of networks of non-volatile memory elements arranged in tiled columns. Logic digitally modulates the equivalent conductance of individual networks among the plurality of networks to map the equivalent conductance of each individual network to a single weight within the neural network. A first partial selection of weights within the neural network is mapped into the equivalent conductances of the networks in the columns to enable the computation of multiply-and-accumulate operations by mixed-signal computation. The logic updates the mappings to select a second partial selection of weights to compute additional multiply-and-accumulate operations and repeats the mapping and computation operations until all computations for the neural network are completed.

Abstract: In one example, an electronic device includes a substrate with a conductive structure and a substrate encapsulant. The conductive structure has a lead with a lead via and a lead protrusion. The lead via can include via lateral sides defined by first concave portions and the lead protrusion can include protrusion lateral sides defined by second concave portions. The substrate encapsulant covers the first concave portions at a first side of the substrate but not the second concave portions so that the lead protrusion protrudes from the substrate encapsulant at a second side of the substrate. An electronic component can be adjacent to the first side of the substrate and electrically coupled to the conductive structure. A body encapsulant encapsulates portions of the electronic component and the substrate. In some examples, the lead can further include a lead trace at the second side of the substrate. In some examples, the substrate can include a redistribution structure at the first side of the substrate.

Abstract: A classification device includes a generation unit, a learning unit, a classification unit, and an output control unit. The generation unit generates pseudo data having a feature similar to a feature of training data. The learning unit learns, by using the training data and the pseudo data, a classification model that classifies data into one of a pseudo class for classifying the pseudo data and a plurality of classification classes other than the pseudo class and that is constructed by a neural network. The classification unit classifies, by using the classification model, input data as a target for classification into one of the pseudo class and the plurality of classification classes. The output control unit outputs information indicating that the input data classified into the pseudo class is data not belonging to any of the plurality of classification classes.