Abstract: An intelligent data and knowledge-driven method for modulation recognition includes the following steps: collecting spectrum data; constructing corresponding attribute vector labels for different modulation schemes; constructing and pre-training an attribute learning model based on the attribute vector labels for different modulation schemes; constructing and pre-training a visual model for modulation recognition; constructing a feature space transformation model, and constructing an intelligent data and knowledge-driven model for modulation recognition based on the attribute learning model and the visual model; transferring parameters of the pre-trained visual model and the pre-trained attribute learning model and retraining the transformation model; and determining whether training on a network is completed and outputting a classification result.

Abstract: A method of training a generator G of a Generative Adversarial Network (GAN) includes receiving, by an encoder E, a target data Y; receiving, by the encoder E, an output G(Z) of the generator G, where the generator G generates the output G(Z) in response to receiving a random sample Z and where a discriminator D of the GAN is trained to distinguish which of the G(Z) and the target data Y; training the encoder E to minimize a difference between a first latent space representation E(G(Z)) of the output G(Z) and a second latent space representation E(Y) of the target data Y, where the output G(Z) and the target data Y are input to the encoder E; and using the first latent space representation E(G(Z)) and the second latent space representation E(Y) to constrain the training of the generator G.

Abstract: An all-optical Diffractive Deep Neural Network (D2NN) architecture learns to implement various functions or tasks after deep learning-based design of the passive diffractive or reflective substrate layers that work collectively to perform the desired function or task. This architecture was successfully confirmed experimentally by creating 3D-printed D2NNs that learned to implement handwritten classifications and lens function at the terahertz spectrum. This all-optical deep learning framework can perform, at the speed of light, various complex functions and tasks that computer-based neural networks can implement, and will find applications in all-optical image analysis, feature detection and object classification, also enabling new camera designs and optical components that can learn to perform unique tasks using D2NNs. In alternative embodiments, the all-optical D2NN is used as a front-end in conjunction with a trained, digital neural network back-end.

Abstract: Certain aspects provide a method, including: receiving a depth image from a depth sensor; receiving a segmentation mask corresponding to the depth image and segmenting the depth image into a set of foreground pixels and a set of background pixels; determining a set of seed pixels in the depth image; for each respective seed pixel of the set of seed pixels: determining a sampling line in the depth image that starts at the respective seed pixel and passes through a portion of the depth image; for each respective sampling line pixel in the sampling line having a value in the segmentation mask indicating a foreground object in the depth image: determining one or more data attribute values based on a depth value for the respective sampling line pixel in the depth image; and adding the one or more data attribute values to a feature vector.

Abstract: A mobile device can implement a neural network-based style transfer scheme to modify an image in a first style to a second style. The style transfer scheme can be configured to detect an object in the image, apply an effect to the image, and blend the image using color space adjustments and blending schemes to generate a realistic result image. The style transfer scheme can further be configured to efficiently execute on the constrained device by removing operational layers based on resources available on the mobile device.

Abstract: Mechanisms are provided to provide an improved computer tool for determining and mitigating the presence of adversarial inputs to an image classification computing model. A machine learning computer model processes input data representing a first image to generate a first classification output. A cohort of second image(s), that are visually similar to the first image, is generated based on a comparison of visual characteristics of the first image to visual characteristics of images in an image repository. A cohort-based machine learning computer model processes the cohort of second image(s) to generate a second classification output and the first classification output is compared to the second classification output to determine if the first image is an adversarial image. In response to the first image being determined to be an adversarial image, a mitigation operation by a mitigation system is initiated.

Abstract: An image of a check is captured by an imaging device and a digital image of the check on a replacement background may be created. The check may be placed on any background while the image of the check is being captured. The replacement background replaces, in the digital image, the background that the check is placed on while its image is being captured. The replacement background may comprise a predetermined image or color(s). The image of the check and the replacement background may be provided into a digital image file that may be transmitted to an institution system for deposit of the check into an account.

Abstract: A method for determining frame timing, a terminal device and a network device are provided. The method for determining frame timing includes that: a terminal device receives beam-specific information sent by a network device through a beam, here, the beam-specific information includes a sequence number of a time-domain location where a synchronization signal is sent through the beam; the terminal device determines a time-domain offset between the synchronization signal and the frame timing according to a correspondence of a sequence number of the beam, the sequence number of the time-domain location where the synchronization signal is sent and the time-domain offset; and the terminal device determines the frame timing according to the time-domain offset.

Abstract: An image processing system accesses an image of a completed form document. The image of the form document includes one or more features, such as form text, at particular locations within the image. The image processing system accesses a template of the form document and computes a rotation and zoom of the image of the form document relative to the template of the form document based on the locations of the features within the image of the form document relative to the locations of the corresponding features within the template of the form document. The image processing system performs a rotation operation and a zoom operation on the image of the form document, and extracts data entered into fields of the modified image of the form document. The extracted data can be then accessed or stored for subsequent use.

Abstract: A method and a system for retrieving video temporal segments are provided. In the method, a video is analyzed to obtain frame feature information of the video; the frame feature information is input into an encoder to output first data relating to temporal information of the video; the first data and a retrieval description for retrieving video temporal segments of the video are input into a decoder to output second data; attention computation training is conducted according to the first data and the second data; video temporal segments of the video corresponding to the retrieval description are determined according to the attention computation training.

Abstract: The present disclosure relates to systems and method for deploying a fiber optic network. Distribution devices are used to index fibers within the system to ensure that live fibers are provided at output locations throughout the system. In an example, fibers can be indexed in multiple directions within the system. In an example, spare ports can be providing in a forward direction and reverse direction ports can also be provided.

Abstract: A method, device and computer program product for deep learning are provided. According to one example, a parameter related to a deep learning model for a training dataset allocated to a server is obtained at a client; a transmission state of the parameter is determined, the transmission state indicating whether the parameter has been transmitted to the server; and information associated with the parameter to be sent to the server is determined based on the transmission state to update the deep learning model. Therefore, the performance of deep learning may be improved, and the network load of deep leaning may be reduced.

Abstract: Provided is a radio communication device which can separate propagation paths of antenna ports and improve a channel estimation accuracy even when using virtual antennas. The device includes: a mapping unit which maps a data signal after modulation to a virtual antenna and a virtual antenna; a phase inversion unit which inverts the phase of S0 transmitted from an antenna port in synchronization with a phase inversion unit between the odd-number slot and the even-number slot; the phase inversion unit which inverts the phase of R0 transmitted from the antenna port; a phase inversion unit which inverts the phase of S1 transmitted from an antenna port in synchronization with a phase inversion unit; and the phase inversion unit which inverts the phase of R1 transmitted from an antenna port.

Abstract: Various face discrimination systems may benefit from techniques for providing increased accuracy. For example, certain discriminative face verification systems can benefit from L2-constrained softmax loss. A method can include applying an image of a face as an input to a deep convolutional neural network. The method can also include applying an output of a fully connected layer of the deep convolutional neural network to an L2-normalizing layer. The method can further include determining softmax loss based on an output of the L2-normalizing layer.

Abstract: The present disclosure relates to methods of analyzing works of art for purposes of authentication or attribution. Such methods may be implemented by receiving digital image data associated with a work of art, identifying a plurality of artist's strokes formed along a surface of the work of art, segmenting the plurality of strokes into a plurality of individual strokes, analyzing the plurality of individual strokes to determine stroke characteristics, and comparing the stroke characteristics to stroke characteristics derived from one or more computational models based on known works of art.

Abstract: An image recognition neural network processing method includes: a compiler segments an image recognition neural network to obtain tiles of at least one network layer group; classifies the tiles of each network layer group; and for each network layer group, generates an assembly code and tile information of the network layer group according to a tile result and a classification result of the network layer group. The same type of tiles correspond to the same assembly function, each assembly code includes a code segment of the assembly function corresponding to each type of tiles, the tile information includes block information of each tile in the network layer group, the tile information used to instruct a neural network processor to, according to the block information therein, invoke a corresponding code segment to process image data of a corresponding tile when a target image is identified by the image recognition neural network.

Abstract: Multiple mobility sets are maintained for nodes of radio networks. The sets comprise information such as: transmit and receive point identities; cell identities; beam identities; frequency channels; channel bandwidth; and black lists. The sets may be defined at different levels, such as network and physical (PHY) level. A network mobility set, e.g., a new-radio (NR) mobility set may, be determined by the gNB, the cell, the UE, or another device. Multiple radio access network nodes and UEs may exchange mobility set information to achieve a distributed mobility solution. A UE may monitor its orientation relative to a TRP, e.g., via use of an onboard MEMS gyroscope, and alter its beamforming parameters in response to changes in orientation and/or changes in TRP connection strength. Cell selection and reselection for beam based networks may use Single Frequency Network (SFN) broadcast of initial access signals without beam sweeping.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a synthesized signal. In some implementations, a computer-implemented system obtains generator input data including at least an input signal having one or more first characteristics, processes the generator input data to generate output data including a synthesized signal having one or more second characteristics using a generator neural network, and outputs the synthesized signal to a device. The generator neural network is trained, based on a plurality of training examples, with a discriminator neural network.

Abstract: A method and a system are provided for training a machine-learning (ML) system to function as a chatbot. According to one embodiment, a method for training and ML system includes providing to the machine-learning system: in a first iteration, a first input-output pair that includes a first input and a first output; and, in a second iteration, a second input-output pair that includes a second input and a second output, where the second input includes the first input-output pair and the second output is different from the first output, so that a context for the second input-output pair is stored in the memory of the ML system.

Abstract: Embodiments for processing data with multiple machine learning models are provided. Input data is received. The input data is caused to be evaluated by a first machine learning model to generate a first inference result. The first inference result is compared to at least one quality of service (QoS) parameter. Based on the comparison of the first inference result to the at least one QoS parameter, the input data is caused to be evaluated by a second machine learning model to generate a second inference result.