Abstract: Concepts and technologies directed to surrogate metadata aggregation for dynamic content assembly are disclosed. Embodiments can include a system that comprises a processor and a memory that stores computer-executable instructions that configure a processor to perform operations. The operations can include obtaining a first visual content from a digital data store, where the first visual content is configured to digitally represent a first scene. The operations can include performing image recognition on the first visual content so as to identify a second visual content that is digitally configured to represent a second scene. The operations can include determining that native original metadata cannot be obtained for the first visual content, where the native original metadata includes information about the first scene digitally represented by the first visual content.

Abstract: A method is disclosed including: receiving raw image data corresponding to a series of raw images; processing the raw image data with an encoder to generate encoded data, where the encoder is characterized by an input/output transformation that substantially mimics the input/output transformation of one or more retinal cells of a vertebrate retina; and applying a first machine vision algorithm to data generated based at least in part on the encoded data.

Abstract: Systems, devices, media, and methods are presented for segmenting an image of a video stream with a client device, extracting one or more color from the image and modifying the video stream. The systems, devices, and method convert images of the set of images to a coordinate representation, perform a histogram equalization, identifies one or more colors of the coordinate representation based on an area of interest, determines a prevailing color, and applies the prevailing color to pixels of the video stream.

Abstract: Data of a captured image including a polarized image is acquired (S10), and space information relating to a position and a posture of a subject in a real space and the position and the posture on an imaging plane is acquired using the captured image data (S12). Next, a polarization degree distribution is acquired from a polarized image of a plurality of orientations (S14), and a position of a light source is acquired by specifying an image of a true light source by threshold value determination of the polarization degree (S16). A reflection characteristic is acquired by applying a rendering equation under assumption that luminance of the captured image is already known (S18), and a material suitable therewith is specified as a material of the subject (S20). Processing according to the material is performed to generate output data and output the data (S22).

Abstract: A computer implemented method to present digital images may include storing a digital image in a database and applying a digital image processing technique to the digital image to identify a region of interest of the digital image. The method may also include storing region data that identifies the region of interest of the digital image in the database and receiving a request for information associated with the digital image from a digital device. In response to the request, the method may include providing the digital image and the region data for transmission to the digital device, the digital device configured to adjust a cropping view of the digital image based on the region data to display the region of interest of the digital image.

Abstract: A glaucoma diagnosis apparatus according to an embodiment includes a fundus image processor configured to receive a fundus image and extract a first region of interest (ROI) and a second ROI from the received fundus image, an image classification neural network configured to learn the extracted first ROI and perform classification into a normal fundus image and a glaucoma fundus image on the basis of the learned first ROI, a vertical cup-to-disc ratio (vCDR) calculator configured to recognize an optic disc (OD) and an optic cup (OC) from the extracted second ROI and calculate a vCDR, and a determinator configured to aggregate a vCDR calculation result and an image classification result of the image classification neural network to determine whether glaucoma is present in the fundus image.

Abstract: A human hairstyle generation method based on multi-feature retrieval and deformation includes: acquiring a hair style mask; identifying feature points of a human face and match the feature points with a hair style database; aligning an image with a standard human face to acquire a corresponding hair region; calculating Minkowski distances between the hair region and hair masks of all frontal faces in the hair style database; assigning corresponding weights after sorting the Minkowski distances from small to large; training a deep learning network to detect hair styles of hair basic blocks at different scales; and taking out a most similar hair style picture. The present invention utilizes a single frontal photo of the human face, and retrieves a three-dimensional hair model most similar to the photo by retrieving a database in a mass three-dimensional hair style database, to avoid manual modeling, thereby improving efficiency and ensures high fidelity.

Abstract: A system and method for classifying a structure or material in an image of a subject. The system comprises: a segmenter configured to form one or more segmentations of a structure or material in an image and generate from the segmentations one or more segmentation maps of the image including categorizations of pixels or voxels of the segmentation maps assigned from one or more respective predefined sets of categories; a classifier that implements a classification machine learning model configured to generate, based on the segmentations maps, one or more classifications and to assign to the classifications respective scores indicative of a likelihood that the structure or material, or the subject, falls into the respective classifications; and an output for outputting a result indicative of the classifications and scores.

Abstract: Embodiments of the disclosure provide a multi-class classification system. An exemplary system includes at least one processor and at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the system to perform operations. The operation includes applying a multi-class classifier to classify a set of objects into multiple classes and applying a plurality of binary classifiers to the set of objects, wherein the plurality of binary classifiers are decomposed from the multi-class classifier, each binary classifier classifying the set of the objects into a first group consisting of one or more classes selected from the multiple classes and a second group consisting of one or more remaining classes of the multiple classes. The operation also includes jointly classifying the set of objects using the multi-class classifier and the plurality of binary classifiers.

Abstract: According to an aspect of an embodiment, operations may comprise obtaining a first point cloud that includes a first point. The operations also comprises obtaining a second point cloud that is a copy of the first point cloud and that includes a second point that is a copy of the first point. The operations also comprises moving the second point cloud with respect to the first point cloud according to a first vector. The operations also comprises identifying a closest point of the first point cloud that is closest to the second point of the second point cloud. The operations also comprises determining a second vector between the closest point and the second point. The operations also comprises determining a measure of usefulness of the first point based on the first vector and the second vector. The operations also comprises indicating the measure of usefulness of the first point.

Abstract: A system and method for detecting a potential match between a candidate facial image and a dataset of facial images is described. Some implementations of the invention determine whether a candidate facial image (or multiple facial images) of a person taken, for example, at point of entry corresponds to one or more facial images stored in a dataset of persons of interest (e.g., suspects, criminals, terrorists, employees, VIPs, “whales,” etc.). Some implementations of the invention detect potential fraud in a dataset of facial images. In a first form of potential fraud, a same facial image is associated with multiple identities. In a second form of potential fraud, different facial images are associated with a single identity, as in the case, for example, of identity theft. According to various implementations of the invention, spectral clustering techniques are used to determine a likelihood that pairs of facial images (or pairs of facial image sets) correspond to the person or different persons.

Abstract: A calculation method using pixel-channel shuffle convolutional neural network is provided. In the method, an operating system receives original input data. The original input data is pre-processed by a pixel shuffle process to be separated into multiple groups in order to minimize dimension of the data. The multiple groups of data are then processed by a channel shuffle process so as to form multiple groups of new input data selected for convolution operation. The unselected data are abandoned. Therefore, the dimension of the input data can be much effectively minimized. A multiplier-accumulator of the operating system is used to execute convolution operation using a convolution kernel and the multiple new groups of input data. Multiple output data are then produced.

Abstract: The technical problem of identifying relevant social proof information with respect to a premium service for a given member profile in an online connection network system is addressed by, first, capturing the associated member's intent based on the member's activity on the web site provided by the online connection network system. The determined intent is used as input into a relevance machine learning model that is executed to identify the member's connection who is a subscriber to the premium service and who has been identified as the most convincing resonated connection of the member with respect to subscribing to the premium service.

Abstract: A method, system and computer-usable medium are disclosed for machine learning to identify service request records associated with an account that is likely to escalate. Certain aspects of the disclosure include generating a random forest model using a training set of service request records to determine a probability of escalation for service requests of the training set; applying the random forest model to a current set of service request records to determine an escalation probability for service requests in the current set; and assigning service request records in the current set to a plurality of escalation probability bins, wherein the service request records of the current set are generally equally divided between the plurality of escalation probability bins, and wherein the service request records of the current set are assigned to a probability bin based on the escalation probability of the service request record.

Abstract: Methods and apparatus are disclosed for implementing machine learning data augmentation within the die of a non-volatile memory (NVM) apparatus using on-chip circuit components formed on or within the die. Some particular aspects relate to configuring under-the-array or next-to-the-array components of the die to generate augmented versions of images for use in training a Deep Learning Accelerator of an image recognition system by rotating, translating, skewing, cropping, etc., a set of initial training images obtained from a host device. Other aspects relate to configuring under-the-array or next-to-the-array components of the die to generate noise-augmented images by, for example, storing and then reading training images from worn regions of a NAND array to inject noise into the images.

Abstract: A method of training an image classification model includes obtaining training images associated with labels, where two or more labels of the labels are associated with each of the training images and where each label of the two or more labels corresponds to an image classification class. The method further includes classifying training images into one or more classes using a deep convolutional neural network, and comparing the classification of the training images against labels associated with the training images. The method also includes updating parameters of the deep convolutional neural network based on the comparison of the classification of the training images against the labels associated with the training images.

Abstract: Embodiments of the present disclosure provide a method for generating an information assessment model, a method for determining the usefulness of comment information, apparatus, electronic device, and computer-readable medium.

Abstract: Remote neural network retraining for surveillance systems is disclosed. In some cases, systems and methods enable remote training and retraining of neural networks, while processing data in real life using the trained and retrained neural networks locally at the surveillance system. The surveillance system can determine a change of location and can retrain the neural network and/or initiate the retraining of the neural network remotely, such as on a cloud server, to retrain the neural network based on new image and/or video data taken from the new location. The remote server can transmit the retrained neural network and/or weights for nodes of the retrained neural network back to the surveillance system. The surveillance system can then update its neural network and process future image and/or video data based on the retrained neural network and/or weights.

Abstract: An image segmentation method system, the system comprising: a training subsystem configured to train a segmentation machine learning model using annotated training data comprising images associated with respective segmentation annotations, so as to generate a trained segmentation machine learning model; a model evaluator; and a segmentation subsystem configured to perform segmentation of a structure or material in an image using the trained segmentation machine learning model. The model evaluator is configured to evaluate the segmentation machine learning model by (i) controlling the segmentation subsystem to segment at least one evaluation image associated with an existing segmentation annotation using the segmentation machine learning model and thereby generate a segmentation of the annotated evaluation image, and (ii) forming a comparison of the segmentation of the annotated evaluation image and the existing segmentation annotation.

Abstract: A processor-implemented classification method includes: determining a first probability vector including a first probability, for each of a plurality of classes, resulting from a classification of an input with respect to the classes; determining, based on the determined first probability vector, whether one or more of the classes represented in the first probability vector are confusing classes; adjusting, in response to one or more of the classes being the confusing classes, the determined first probability vector based on a first probability of each of the confusing classes and a maximum value of the first probabilities; determining a second probability vector including a second probability, for each of the classes, resulting from another classification of the input with respect to the classes; and performing classification on the input based on a result of a comparison between the determined second probability vector and the adjusted first probability vector.

Abstract: Disclosed is an electronic apparatus comprising, a memory configured to store instructions; and at least one processor connected to the memory, and configured to detect at least one object of a first-class object or a second-class object included in a target image by the electronic apparatus using an artificial intelligent algorithm to apply the target image to a learned neural network model, and identify and apply an image-quality processing method to be individually applied to at least one detected object, the neural network model is set to detect an object included in an image, as trained based on learning data such as an image, a class to which the image belongs, information about the first-class object included in the image, and information about the second-class object included in the image.

Abstract: An image processing method and apparatus, and a storage medium are provided. The method includes: obtaining a first image and a second image of a to-be-authenticated object, where the first image is captured by a first camera module, and the second image is captured by at least one second camera module; comparing the first image with image data in a target library for identity authentication, to obtain a first authentication result; and in response to that the first authentication result is authentication failure, performing joint authentication on the first image and the second image, and determining the identity of the to-be-authenticated object according to a second authentication result of the joint authentication.

Abstract: This application provides an image recognition method and device. The image recognition method includes: setting a presentation time sequence corresponding to an image sequence includes N images, the presentation time sequence includes unequal presentation times, a difference between any two presentation times of the unequal presentation times is k×?, k is a positive integer, and ? is a preset time period value; processing the image sequence by using a computer vision algorithm, to obtain a computer vision signal corresponding to each image in the image sequence; obtaining a feedback signal that is corresponding to each image in the image sequence generated when an observation object watches the image sequence displayed in the presentation time sequence; and fusing, for each image in the image sequence, a corresponding computer vision signal and a corresponding feedback signal to obtain a target recognition signal of each image in the image sequence.

Abstract: Visibility of a license plate and color reproducibility of a vehicle body are improved in a monitoring camera. A vehicle body area detection unit detects a vehicle body area of a vehicle from an image signal. A license plate area detection unit detects a license plate area of the vehicle from the image signal. A vehicle body area image processing unit performs processing of the image signal corresponding to the detected vehicle body area. A license plate area image processing unit performs processing different from the processing of the image signal corresponding to the vehicle body area on the image signal corresponding to the detected license plate area. A synthesis unit synthesizes the processed image signal corresponding to the vehicle body area and the processed image signal corresponding to the license plate area.

Abstract: A method of optimizing printing is applied to a printing apparatus and has following steps of retrieving an input image, executing a tagging process on the input image for tagging one of an edge tag, a fuzzy tag, and a photo tag on each sub-image of the input image, executing the different printing converting process on each sub-image according to its tag for obtaining a printable image, and printing according to the printable image. The present disclosed example can deepen the object edges in the image being printed, improving the image quality of non-edge regions, and improving the printing quality.

Abstract: An equipment management system is provided with an RFID provided for each equipment, an equipment position information storage unit, an equipment management data storage unit, an image information storage unit, a reader/writer, and a portable information terminal transmitting and receiving information with each storage unit. The image information storage unit has point group information and other image information of the inside of a facility. The equipment position information storage unit has installation position information of the equipment associated with the RFID and a virtual tag allowing reference to equipment management data on the equipment management data storage unit corresponding to the equipment and the image information associated with the equipment associated with the RFID in the image information storage unit.

Abstract: According to one embodiment, a barcode generation device includes an interface and a processor. The interface receives a barcode display instruction. The processor generates a basic barcode in response to the display instruction received by the interface, generates a plurality of barcodes in which noise is superimposed at different positions on the basic barcode, determines whether there is misreading in an image scanned by a barcode reader on a screen on which the plurality of barcodes are sequentially displayed, and sequentially displays the plurality of barcodes on a display device when it is determined that there is no misreading.

Abstract: A transaction card may comprise a first card component having an electrically conductive surface configured to receive an electrically applied coating. An electrically applied coating may be formed on the electrically conductive surface. The transaction card may be manufactured by forming a first card component having an electrically conductive surface configured to receive an electrically applied coating. The method may also include applying an electrically applied coating to the electrically conductive surface.

Abstract: An account is managed using information read from a dual frequency transponder. Information stored on the dual frequency transponder can be read by a NFC-enabled device and by a UHF RFID reader. The information links, corresponds, or otherwise provides access to account information stored at a remote server. For example, a NFC-enabled device can read the information from the dual frequency transponder and use that information to enable instant and on-the-spot recharging of a toll account. In addition, a UHF RFID toll reader can scan information from the dual frequency transponder and use that information to debit toll charges from the correct toll account. The dual frequency transponder can be embedded in a license plate and read using a reader placed in the road. Additionally, the transponder can be configured to function at the correct frequency only when a valid vehicle registration sticker is applied to the license plate.

Abstract: An industrial system includes an electronic badge worn or otherwise transported by an industrial vehicle operator. The electronic badge has a housing, a processor, and a transceiver coupled to the processor that communicates on a personal-area network with a badge communicator that is provided on an industrial vehicle when the electronic badge and the badge communicator are in range of each other. Further, an activity sensor collects activity information about the industrial vehicle operator as the industrial vehicle operator performs work tasks. The electronic badge exchanges data collected by the activity sensor with the industrial vehicle, for communication to a remote server. An electronic message is communicated to the industrial vehicle for output to a display thereon, and the electronic message defines an assigned task that is based upon previously collected data from the activity sensor.

Abstract: Briefly, a method for verifying the visual perceptibility of a display is provided. An intended message is written to a bistable display. Pixels that comprise portions of the message are measured and evaluated to determine if the message actually displayed on the bistable display was perceptible by a human or a machine. In some cases, information regarding the message actually perceivable from the display may be stored for later use. Responsive to determining that a message is perceivable or not perceivable, alarms may be set, one or more third parties notified, or additional display features may be set.

Abstract: A packaging system for at least one article includes a multilayer laminate structure that encapsulates a transformative material between first and second conductive layers where one of the first and second conductive layers defines a set of elongate sections. An NFC/RFID IC is electrically coupled to an antenna. The NFC/RFID IC has a plurality of input terminals electrically coupled to a plurality of electrical circuits that provide for electrical connection or electrical disconnection between sections in the set of elongate sections in accordance with a predefined codeword. The NFC/RFID IC is configured to sense voltage signals produced by the plurality of electrical circuits, determine a sensed codeword based on the sensed voltage signals, compare the sensed codeword to the predefined codeword, and output a signal based on such comparison.

Abstract: Example apparatus and methods for generating context-aware artificial intelligence characters are disclosed. An example apparatus to animate an artificial intelligence character includes a data tagger to tag data in a media data stream to generate a plurality of data files of tagged data, the data files corresponding to different time periods in a storyline, the tagged data associated with a first character in the media data stream, the artificial intelligence character to portray the first character. The example apparatus includes a trainer to generate a response model of the first character based on the data file corresponding to a current data time period and one or more data files corresponding to one or more earlier time periods of the storyline and a response generator to apply the response model based on a stimulus input to animate the artificial intelligence character.

Abstract: A system and method for controlling a nodal network. The method includes estimating an effect on the objective caused by the existence or non-existence of a direct connection between a pair of nodes and changing a structure of the nodal network based at least in part on the estimate of the effect. A nodal network includes a strict partially ordered set, a weighted directed acyclic graph, an artificial neural network, and/or a layered feed-forward neural network.

Abstract: Apparatuses and methods of manufacturing same, systems, and methods for generating a convolutional neural network (CNN) are described. In one aspect, a minimal CNN having, e.g., three or more layers is trained. Cascade training may be performed on the trained CNN to insert one or more intermediate layers until a training error is less than a threshold. When cascade training is complete, cascade network trimming of the CNN output from the cascade training may be performed to improve computational efficiency. To further reduce network parameters, convolutional filters may be replaced with dilated convolutional filters with the same receptive field, followed by additional training/fine-tuning.

Abstract: An object detection system includes an image capture device, a memory, and a processor. The image capture device captures an image. The memory stores an instruction corresponding to an inference engine based on a multi-scale convolutional neural network architecture including a first, a second, and an object detection scale. The processor executes the instruction to: reduce network widths of convolution layers of the second scale; run the inference engine according to the adjusted convolutional neural network architecture to receive the image as an initial input; input a first output generated by the first scale according to the initial input into the second and the object detection scale; input a second output generated by the second scale according to the first output into the object detection scale; generate a final output according to the first and the second output by the object detection scale, to perform object detection on the image.

Abstract: A new approach is proposed to support efficient convolution for deep learning by vectorizing multi-dimensional input data for multi-dimensional fast Fourier transform (FFT) and direct memory access (DMA) for data transfer. Specifically, a deep learning processor (DLP) includes a plurality of tensor engines each configured to perform convolution operations by applying one or more kernels on multi-dimensional input data for pattern recognition and classification based on a neural network, wherein each tensor engine includes, among other components, one or more vector processing engines each configured to vectorize the multi-dimensional input data at each layer of the neural network to generate a plurality of vectors and to perform multi-dimensional FFT on the generated vectors and/or the kernels to create output for the convolution operations. Each tensor engine further includes a data engine configured to prefetch the multi-dimensional data and/or the kernels to both on-chip and external memories via DMA.

Abstract: A hardware architecture and a processing method for an activation function in a neural network are provided. A look-up table, which is a corresponding relation among multiple input ranges and linear functions, is provided. A difference between an initial value and an end value of the input range of each linear function is an exponentiation of base-2. These linear functions form a piecewise linear function to approximate the activation function. At least one bit value of an input value is used as an index to query the look-up table to determine a corresponding linear function. The part of bits value of the input value is fed into the determined linear function to obtain an output value. Accordingly, a range comparison may be omitted, and the number of bits of a multiplier-accumulator may be reduced, so as to achieve the objectives of low costs and low power consumption.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating output examples using neural networks. One of the methods includes, at each generation time step, processing a first recurrent input comprising an N-bit output value at the preceding generation time step in the sequence using a recurrent neural network and in accordance with a hidden state to generate a first score distribution; selecting, using the first score distribution, values for the first half of the N bits; processing a second recurrent input comprising (i) the N-bit output value at the preceding generation time step and (ii) the values for the first half of the N bits using the recurrent neural network and in accordance with the same hidden state to generate a second score distribution; and selecting, using the second score distribution, values for the second half of the N bits of the output value.

Abstract: An integrated circuit chip apparatus and a processing method performed by an integrated circuit chip apparatus are disclosed. The disclosed integrated circuit chip apparatus and processing method are used for executing a multiplication operation, a convolution operation, or a training operation of a neural network. The present technical solution has the advantages of a reduced computational cost and low power consumption.

Abstract: A spiking neural network includes a plurality of neurons implemented in respective circuits. Each neuron produces a continuous-valued membrane potential according to a Growth Transform bounded by an extrinsic energy constraint. The continuous-valued membrane potential is defined as a function of spiking current received from another neuron in the plurality of neurons, and a received electrical current stimulus. The spiking neural network includes a network energy function representing network energy consumed by the plurality of neurons and a neuromorphic framework.

Abstract: Disclosed herein are system, method, and computer program product embodiments for classifying data objects using machine learning. In an embodiment, an artificial neural network may be trained to identify explained variable values corresponding to data object attributes. For example, the explained variables may be a category and a subcategory with the subcategory having a hierarchical relationship to the category. The artificial neural network may then receive a data record having one or more attribute values. The neural network may then identify a first and second explained variable value corresponding to the one or more attribute values based on the trained neural network model. The first and second explained variable values may then be associated with the data record. For example, if the data record is stored in a database, the record may be updated to include the first and second explained variable values.

Abstract: In an approach to determining the effectiveness of a proposed solution, one or more computer processors monitor real-time communications. The one or more computer processors identify or more topics associated with the monitored real-time communications. The one or more computer processors feed the identified one or more topics and associated real-time communications into a solution efficacy model. The one or more computer processors generate based on one or more calculations by the solution efficacy model, an efficacy rating for the identified real-time communications. The one or more computer processors generate a prioritization of the identified real-time communications based on the generated efficacy rating.

Abstract: A system and method for predicting whether experimental legislation will become enacted into law may include memory and at least one processor configured to receive a first request which proposes test content as an experimental law and receive a second request pertaining to a selected number of sponsors. The processor may automatically import, over a network, data from databases, prepare the data as predictive modeling data, split the predictive modeling data into two sets of data, train a two-class neural network on training data to predict whether the test law will become law, generate a set of results from the training data, cross-validate the set of results with the test data, and deploy, over the network, a predictive performance.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media for generating a recommended target audience based on determining a predicted attendance utilizing a neural network approach. For example, the disclosed systems can utilize an approximate nearest neighbor algorithm to identify individuals that are within a similarity threshold of invitees for an event. In addition, the disclosed systems can implement an attendance prediction model to determine a probability of an invitee attending the event. The disclosed systems can further determine a predicted attendance for an event based on the individual probabilities. Based on identifying the similar individuals to, and the attendance probabilities for, the invitees, the disclosed systems can generate a recommended target audience to satisfy a target attendance for an event based on a predicted attendance for the event.

Abstract: A method is provided, including: storing comments generated in response to a content item served over a network; analyzing the comments to determine features associated with each of the comments; using a scoring model to score each comment based on the comment's corresponding features; receiving a request to serve a subset of the comments; responsive to the request, selecting a ranking of the comments that is one permutation from possible rankings of the comments, wherein selecting the ranking is in accordance with a probability distribution of the possible rankings that is based on the scores of the comments; serving comments identified by the selected ranking over the network to a client device; determining a dwell time on the served comments; applying the dwell time to update the scoring model.

Abstract: An example method of training a neural network includes defining hardware building blocks (HBBs), neuron equivalents (NEQs), and conversion procedures from NEQs to HBBs; defining the neural network using the NEQs in a machine learning framework; training the neural network on a training platform; and converting the neural network as trained into a netlist of HBBs using the conversion procedures to convert the NEQs in the neural network to the HBBs of the netlist.

Abstract: Implementations are described herein for training and applying machine learning models to digital images capturing plants, and to other data indicative of attributes of individual plants captured in the digital images, to recognize individual plants in distinction from other individual plants. In various implementations, a digital image that captures a first plant of a plurality of plants may be applied, along with additional data indicative of an additional attribute of the first plant observed when the digital image was taken, as input across a machine learning model to generate output. Based on the output, an association may be stored in memory, e.g., of a database, between the digital image that captures the first plant and one or more previously-captured digital images of the first plant.

Abstract: Batch normalization (BN) layer fusion and quantization method for model inference in artificial intelligence (AI) network engine are disclosed. A method for a neural network (NN) includes merging batch normalization (BN) layer parameters with NN layer parameters and computing merged BN layer and NN layer functions using the merged BN and NN layer parameters. A rectified linear unit (RELU) function can be merged with the BN and NN layer functions.

Abstract: In an approach to determining the effectiveness of a proposed solution, one or more computer processors monitor real-time communications. The one or more computer processors identify or more topics associated with the monitored real-time communications. The one or more computer processors feed the identified one or more topics and associated real-time communications into a solution efficacy model. The one or more computer processors generate based on one or more calculations by the solution efficacy model, an efficacy rating for the identified real-time communications. The one or more computer processors generate a prioritization of the identified real-time communications based on the generated efficacy rating.