Abstract: Techniques and mechanisms for performing a Bayesian inference with a spiking neural network. In an embodiment, a parent node of the spiking neural network receives a first bias signal which is periodic. The parent node communicates a likelihood signal to a child node, wherein the parent node and the child node correspond to a first condition and a second condition, respectively. Based on a phase change which is applied to the first bias signal, the likelihood signal indicates a probability of the first condition. The child node also receives a signal which indicates an instance of the second condition. Based on the indication and a second bias signal, the child node signals to the first node that an adjustment is to be made to the phase change applied to the first bias signal. After the adjustment, the likelihood signal indicates an updated probability of the first condition.

Abstract: Disclosed herein are systems and methods capable of establishing a communication session between a customer and an analyst. The contents of the communication session are analyzed to make recommendations of goods and services to the customer. Otherwise, the communication session may be redirected from one channel, for example, chatting to another channel, for example, voice call, to another analyst. The customer's information and the communication session details are retained, and provided to another analyst before the customer is redirected. Such systems, apparatuses, methods, and computer program products use real-time machine learning scoring algorithm to determine which analyst the customers should be transferred to, and thereby saving a lot of time for both the analysts and the customers, and significantly reduce misroutes by eliminating human errors.

Abstract: A method may include: analyzing each of a group of inorganic particles to generate data about physicochemical properties of each of the inorganic particles; and generating a correlation between a reactivity index of each of the inorganic particles and the data.

Abstract: Systems and methods for detecting aortic aneurysms using ensemble based deep learning techniques that utilize numerous computed tomography (CT) scans collected from numerous de-identified patients in a database. The system includes software that automates the analysis of a series of CT scans as input (in DICOM file format) and provides output in two dimensions: (1) ranking CT scans by risks of adverse events from aortic aneurysm, (2) providing aortic aneurysm size estimates. A repository of CT scans may be used for training of deep neural networks and additional data may be drawn from localized patient information from institutions and hospitals which grant permission.

Abstract: A recommendation system for recommending a target feature value for a target feature for a target deployment is provided. The recommendation system, for each of a plurality of deployments, collects feature values for the features of that deployment. The recommendation system then generates a model for recommending a target feature value for the target feature based on the collected feature values of the features for the deployments. The recommendation system applies the model to the features of the target deployment to identify a target feature value for the target feature. The recommendation system then provides the identified target feature value as a recommendation for the target feature for the target deployment.

Abstract: A wireless network can generate candidate signal configurations for physical transmissions to or from a user equipment (UE) in a radio environment. The generation of candidate signal configurations can be performed using a first neural network that is associated with the UE. These signal configurations can then be evaluated using a second neural network that is associated with the radio environment. The second neural network can be trained using measurements from previous physical transmissions in the radio environment. The trained second neural network generates a reward value that is associated with the candidate signal configurations. The first neural network is then trained using the reward values from the second neural network to produce improved candidate signal configurations. When a signal configuration that produces a suitable reward value is generated, this signal configuration can be used for the physical transmission in the radio environment.

Abstract: A method for detecting defects in a semiconductor device includes pre-training a pre-trained convolutional neural network (CNN) model using a sampled clean data set extracted from a first data set; training a normal convolutional neural network model and a label-noise convolutional neural network model using first data of the first data set and the pre-trained convolutional neural network model. The method also includes outputting a first prediction result on whether second data of a second data set is good or bad using the second data and the normal convolutional neural network model; and outputting a second prediction result on whether second data is good or bad using the second data and the label-noise convolutional neural network model. The first prediction result is compared with the second prediction result to perform noise correction when there is a label difference. Third data created as results of the noise correction is added to the sampled clean data set.

Abstract: Video processing with a multi-quality loop filter using a multi-task neural network is performed by at least one processor and includes generating model IDs, based on quantization parameters in an input, selecting a first set of masks, each mask in the first set of masks corresponding to one of the generated model IDs, performing convolution of first weights of a first set of neural network layers and the selected first set of masks to obtain first masked weights, and selecting a second set of neural network layers and second weights, based on the quantization parameters, generating a quantization parameter value, based on the quantization parameters, and computing an inference output, based on the first masked weights and the second weights, using the generated quantization parameter value.

Abstract: A method of estimating a toggle count of a circuit, includes, in part, simulating the circuit to generate training data and an associated training toggle count of the internal nodes of the circuit in response to a test bench, training a neural network system to generate an estimate of the training toggle count in accordance with the training data and the associated training toggle count, simulating the circuit to generate simulation data in response to a first set of input values applied to the circuit, and invoking the trained neural network system to estimate a number of toggles of the internal nodes of the circuit from the simulation data. The training data may include, in part, values of input signals applied to the circuit and values of registers disposed in the circuit for a multitude of time stamps. The neural network system may include, in part, at least three layers.

Abstract: A hybrid quantum classical (HQC) computer, which includes both a classical computer component and a quantum computer component, implements improvements to the quantum approximate optimization algorithm (QAOA) which enable QAOA to be applied to valuable problem instances (e.g., those including several thousand or more qubits) using near-term quantum computers.

Abstract: A method for detecting fraudulent interactions may include receiving interaction data, including a first plurality of interactions with (first) fraud labels and a second plurality of interactions (without fraud labels). Second fraud label data for each of the second plurality of interactions may be generated with a first neural network (e.g., classifying whether each interaction is fraudulent or not). Generated interaction data and generated fraud label data may be generated with a second neural network. Discrimination data for each of the second plurality of interactions and generated interactions may be generated with a third neural network (e.g., classifying whether the respective interaction is real or not). Error data may be determined based on the discrimination data (e.g., whether the respective interaction is correctly classified). At least one of the neural networks may be trained based on the error data. A system and computer program product are also disclosed.

Abstract: Systems and methods are provided for automatedly troubleshooting a computing application (e.g., a cloud-based computing application). An application domain of the computing application is modeled as a two-dimensional array of cells, a first dimension of the array representing components or microservices of the application domain, and a second dimension of the array representing states of the components or microservices, the array including paths between pairs of cells in the array. A troubleshooting goal is defined as a target state of the application domain, the target state corresponding to a target cell in the array. An initial state of the application domain is also provided, the initial state corresponding to an initial cell in the array. A reinforcement-learning-trained machine-learning algorithm can determine a solution path in the array between the initial cell and the target cell. Divergence between a failure case and a solution path indicates a probable failure cause.

Abstract: A robot system includes a holding compartment configured to transport one or more items and an electronic control unit having a processor and a non-transitory computer readable memory including a machine-readable instruction set. The robot system further includes a camera for capturing image data of an environment of the holding compartment and a robotic arm each communicatively coupled to the electronic control unit. The machine-readable instruction set causes the processor to receive image data of the environment of the holding compartment from the camera, determine a set of handling instructions for an item collected by the robotic arm, determine a location within the holding compartment for placing the item collected by the robotic arm based on the image data of the environment of the holding compartment and the set of handling instructions, and manipulate the robotic arm to place the item within the holding compartment at the determined location.

Abstract: Technical solutions are described for predicting linepack delays. An example method includes receiving temporal sensor measurements of a first fluid-delivery pipeline network and generating a causality graph of the first fluid-delivery pipeline network. The method also includes determining a topological network of the stations based on the causality graph, where the topological network identifies a temporal delay between a pair of stations. The method also includes generating a temporal delay prediction model based on the topological network and predicting the linepack delays of a second fluid-delivery pipeline network based on the temporal delay prediction model, where a compressor station of the second fluid-delivery pipeline network compresses fluid based on the predicted linepack delays to maintain a predetermined pressure.

Abstract: A computer implemented method for adapting a model for recognition processing to a target-domain is disclosed. The method includes preparing a first distribution in relation to a part of the model, in which the first distribution is derived from data of a training-domain for the model. The method also includes obtaining a second distribution in relation to the part of the model by using data of the target-domain. The method further includes tuning one or more parameters of the part of the model so that difference between the first and the second distributions becomes small.

Abstract: Making failure scenarios using adversarial reinforcement learning is performed by storing, in a first storage, a variety of first experiences of failures of a player agent due to an adversarial agent, and performing a simulation of an environment including the player agent and the adversarial agent. It also includes calculating a similarity of a second experience of a failure of the player agent in the simulation and each of the variety of first experiences in the first storage, and updating the first storage by adding the second experience as a new first experience of the variety of first experiences in response to the similarity being less than a threshold. Additionally, the use of adversarial reinforcement learning can include training the adversarial agent by using at least one of the plurality of first experiences in the first storage to generate an adversarial agent having diverse experiences.

Abstract: The system obtains a set of tickets representing customer requests for a customer-support ticketing system. Next, the system produces a set of request vectors by feeding words from each ticket through a model to generate a request vector for the ticket, wherein the set of request vectors is represented as a set of points in a vector space. The system then performs a clustering operation on the set of points to form clusters representing support topics, wherein the clustering operation creates a new point for a new ticket in the vector space when the new ticket is received, and incrementally updates existing clusters to accommodate the new point. Finally, the system presents a user interface to a customer-support agent, wherein the user interface uses the support topics to organize the customer requests, and enables the customer-support agent to perform a customer-support operation in response to a customer request.

Abstract: Provided is a method of adaptively performing artificial intelligence (AI) downscaling on an image during a video telephone call of a user terminal. The method includes obtaining, from an opposite user terminal, AI upscaling support information of the opposite user terminal that is a target of a video telephone call, determining whether the user terminal is to perform AI downscaling on an original image, based on the AI upscaling support information, based on determining that the user terminal is to perform AI downscaling on the original image, obtaining a first image by AI downscaling the original image using a downscaling deep neural network (DNN), generating image data by performing first encoding on the first image, and transmitting AI data including information related to the AI downscaling and the image data.

Abstract: A method including receiving, in a machine learning model (MLM), a corpus including words. The MLM includes layers configured to extract keywords from the corpus, plus a retrospective layer. A first keyword and a second keyword from the corpus are identified in the layers. The first and second keywords are assigned first and second probabilities. Each probability is a likelihood that a keyword is to be included in a key phrase. A determination is made, in the retrospective layer, of a first probability modifier that modifies the first probability based on a first dependence relationship between the second keyword being placed after the first keyword. The first probability is modified using the first probability modifier. The first modified probability is used to determine whether the first keyword and the second keyword together form the key phrase. The key phrase is stored in a non-transitory computer readable storage medium.

Abstract: Techniques for model adaptation are described. For example, a method of receiving a call to provide either a model variant or a model variant profile of a deep learning model, the call including desired performance of the deep learning model, a deep learning model identifier, and current edge device characteristics; comparing the received current edge device characteristics to available model variants and profiles based on the desired performance of the deep learning model to generate or select a model variant or profile, the available model variants and profiles determined by the model identifier; and sending the generated or selected model variant or profile to the edge device to use in inference is detailed.

Abstract: A server inputs behavior information of a target to a trained machine learning model that learn a plurality of association relations obtained by associating combinations of behaviors generated from a plurality of behaviors included in a plurality of training data items with likelihoods indicating certainties that the combinations of the behaviors are in a specific state, the trained machine learning having been trained by using the plurality of training data items obtained by associating combinations of behaviors of persons corresponding to the specific state. The server specifies a difference between the combination of the behaviors in each of the plurality of association relations and the behavior information of the target based on output results of the trained machine learning model, and determines an additional behavior for causing the target to transit to the specific state based on the likelihood the difference.

Abstract: Methods, systems and computer program products for updating a word embedding model are provided. Aspects include receiving a first data set comprising a relational database having a plurality of words. Aspects also include generating a word embedding model comprising a plurality of word vectors by training a neural network using unsupervised machine learning based on the first data set. Each word vector of the plurality of word vector corresponds to a unique word of the plurality of words. Aspects also include storing the plurality of word vectors and a representation of a hidden layer of the neural network. Aspects also include receiving a second data set comprising data that has been added to the relational database. Aspects also include updating the word embedding model based on the second data set and the stored representation of the hidden layer of the neural network.

Abstract: A microprocessor system comprises a computational array and a vector computational unit. The computational array includes a plurality of computation units. The vector computational unit is in communication with the computational array and includes a plurality of processing elements. The processing elements are configured to receive output data elements from the computational array and process in parallel the received output data elements.

Abstract: A processor-implemented verification method includes: detecting a characteristic of an input image; acquiring input feature transformation data and enrolled feature transformation data by respectively transforming input feature data and enrolled feature data based on the detected characteristic, wherein the input feature data is extracted from the input image using a feature extraction model; and verifying a user corresponding to the input image based on a result of comparison between the input feature transformation data and the enrolled feature transformation data.

Abstract: Methods and computer systems improve a trained base deep neural network by structurally changing the base deep neural network to create an updated deep neural network, such that the updated deep neural network has no degradation in performance relative to the base deep neural network on the training data. The updated deep neural network is subsequently training. Also, an asynchronous agent for use in a machine learning system comprises a second machine learning system ML2 that is to be trained to perform some machine learning task. The asynchronous agent further comprises a learning coach LC and an optional data selector machine learning system DS. The purpose of the data selection machine learning system DS is to make the second stage machine learning system ML2 more efficient in its learning (by selecting a set of training data that is smaller but sufficient) and/or more effective (by selecting a set of training data that is focused on an important task).

Abstract: A method may include: providing a model of cement induction time; designing a cement composition, based at least partially, on the model of cement induction time; and preparing the cement composition.

Abstract: Systems and methods for automatic generating of a Bayes net content graph are disclosed herein. The system can include a memory including a mapping matrix. The system can include at least one server. The at least one server can generate a user matrix having n columns and p rows. In some aspects, each of the n columns is associated with a student and each of the p rows is associated with a content item. The at least one server can: store the user matrix in the memory; retrieve the mapping matrix from the memory; iteratively identify prerequisite relationships between the skills identified in the user matrix; generate edges between the skills in the user matrix based on the iteratively identified prerequisite relationships; and orient the edges between the skill.

Abstract: A method includes using a computational network to learn and predict time-series data. The computational network includes one or more layers, each having an encoder and a decoder. The encoder of each layer multiplicatively combines (i) current feed-forward information from a lower layer or a computational network input and (ii) past feedback information from a higher layer or that layer. The encoder of each layer generates current feed-forward information for the higher layer or that layer. The decoder of each layer multiplicatively combines (i) current feedback information from the higher layer or that layer and (ii) at least one of the current feed-forward information from the lower layer or the computational network input or past feed-forward information from the lower layer or the computational network input. The decoder of each layer generates current feedback information for the lower layer or a computational network output.

Abstract: Rule determination for black-box machine-learning models (BBMLMs) is described. These rules are determined by an interpretation system to describe operation of a BBMLM to associate inputs to the BBMLM with observed outputs of the BBMLM and without knowledge of the logic used in operation by the BBMLM to make these associations. To determine these rules, the interpretation system initially generates a proxy black-box model to imitate the behavior of the BBMLM based solely on data indicative of the inputs and observed outputs—since the logic actually used is not available to the system. The interpretation system generates rules describing the operation of the BBMLM by combining conditions—identified based on output of the proxy black-box model—using a genetic algorithm. These rules are output as if-then statements configured with an if-portion formed as a list of the conditions and a then-portion having an indication of the associated observed output.

Abstract: The present disclosure relates to a method of generating spikes by a neuron of a spiking neural network. The method comprises generating at each time, wherein the spike generation encodes at each time instant at least two variable values at the neuron. Synaptic weights may be optimized for a spike train generated by a given presynaptic neuron of a spiking neural network, wherein the spike train being indicative of features of at least one timescale.

Abstract: A method of testing a communication network. The method comprises providing network communication service to communication service subscribers by a first set of computing resources that are part of a communication network, replicating communication data packets, directing the replicated communication data packets to the first set of computing resources and to a second set of computing resources that are part of the communication network but which do not provide network communication service to communication service subscribers, introducing a random error into the second set of computing resources, capturing by a testing application executing on a server computer the outputs from the second set of computing resources after processing the replicated communication data packets in the context of the random error, analyzing the captured outputs by the testing application, determining that the captured outputs indicate a failure by the testing application, and taking action by the testing application.

Abstract: An embodiment includes duplicating an input dataset being input to a model predictive control (MPC) module for input to a first Hierarchical Temporal Memory (HTM) network. The embodiment also includes generating system behavior data using the MPC module for characteristic data of the input dataset. The embodiment also includes generating first HTM prediction data from the input dataset and the system behavior data using the first HTM network, the first HTM prediction data comprising predictions for respective dimensions of the system. The embodiment also includes generating second HTM prediction data from the first HTM prediction data and system output data using a second HTM network, the second HTM prediction data comprising a distinction between the first HTM prediction and the system output data. Finally, the embodiment includes determining that the distinction of the second HTM prediction data indicates an anomaly and adjusting system input data based on the anomaly.

Abstract: The disclosure provides a method for predicting a traffic matrix, a computing device, and a storage medium. The method includes: establishing a dataset based on continuous historical traffic matrices; and inputting one or more historical traffic matrices in the dataset into a trained model for predicting traffic matrices, to obtain one or more predicted traffic matrices. The trained model for predicting traffic matrices is obtained by the following actions: establishing a model for predicting traffic matrices based on a correlation-modeling neural network and a temporal-modeling neural network; and training the model for predicting traffic matrices based on a set of training samples, in which the set of training samples includes sample traffic matrices and label traffic matrices corresponding to the sample traffic matrices at prediction moment samples.

Abstract: Systems, methods, and computer-readable media are disclosed for parallel stochastic gradient descent using linear and non-linear activation functions. One method includes: receiving a set of input examples; receiving a global model; and learning a new global model based on the global model and the set of input examples by iteratively performing the following steps: computing a plurality of local models having a plurality of model parameters based on the global model and at least a portion of the set of input examples; computing, for each local model, a corresponding model combiner based on the global model and at least a portion of the set of input examples; and combining the plurality of local models into the new global model based on the current global model and the plurality of corresponding model combiners.

Abstract: Methods and computer systems improve a trained base deep neural network by structurally changing the base deep neural network to create an updated deep neural network, such that the updated deep neural network has no degradation in performance relative to the base deep neural network on the training data. The updated deep neural network is subsequently training. Also, an asynchronous agent for use in a machine learning system comprises a second machine learning system ML2 that is to be trained to perform some machine learning task. The asynchronous agent further comprises a learning coach LC and an optional data selector machine learning system DS. The purpose of the data selection machine learning system DS is to make the second stage machine learning system ML2 more efficient in its learning (by selecting a set of training data that is smaller but sufficient) and/or more effective (by selecting a set of training data that is focused on an important task).

Abstract: Aspects of the subject disclosure may include, for example, a device configured for generating a number of content groups by grouping a number of tagged media segments according to their corresponding content designators. The device can be further configured for detect an activity of a user, determining a particular segment length and content designator according to the detected activity, and generating a selected content group according to the particular segment length and content designator. A content group is selected according to the particular segment length and the content designator. Other embodiments are disclosed.

Abstract: A robot system includes a robot, a robot controller, a video acquisition device configured to acquire a real video of a work space, and a head-mounted type video display device provided with a visual line tracking section configured to acquire visual line information. A robot controller includes an information storage section configured to store information used for controlling the robot while associating the information with a type of an object, a gaze target identification section configured to identify, in the video, a gaze target viewed by a wearer based on the visual line information, and a display processing section configured to cause the video display device to display the information associated with the object corresponding to the identified gaze target, side by side with the gaze target in the form of one image through which the wearer can visually grasp, select, or set contents of the information.

Abstract: Computer-implemented, machine-learning systems and methods relate to a combination of neural networks. The systems and methods train the respective member networks both (i) to be diverse and yet (ii) according to a common, overall objective. Each member network is trained or retrained jointly with all the other member networks, including member networks that may not have been present in the ensemble when a member is first trained.

Abstract: Provided are a method of generating a trained third neural network to recognize a speaker of a noisy speech signal by combining a trained first neural network which is a skip connection-based neural network for removing noise from the noisy speech signal with a trained second neural network for recognizing the speaker of a speech signal, and a neural network device for operating the neural networks.

Abstract: A prediction system and method may include receiving a plurality of discrete applicant data inputs and a supporting document, the applicant data inputs and the supporting document being relevant to a permit application, providing a first predicted probability of approval of the permit application by comparing the discrete applicant data inputs with weighted criteria of previous applicant profiles stored in a first database, analyzing the supporting document to determine a second predicted probability of approval of the permit application by comparing the supporting document with previous applicant supporting documents stored in a second database, performing a sentiment analysis on external publically available information relevant to at least one aspect of the permit application to determine an impact score on the permit application, and determining an overall probability of success based on the first predicted probability, the second predicted probability, and the impact score.

Abstract: A learning computer system may include a data processing system and a hardware processor and may estimate parameters and states of a stochastic or uncertain system. The system may receive data from a user or other source. Parameters and states of the stochastic or uncertain system are estimated using the received data, numerical perturbations, and previous parameters and states of the stochastic or uncertain system. It is determined whether the generated numerical perturbations satisfy a condition. If the numerical perturbations satisfy the condition, the numerical perturbations are injected into the estimated parameters or states, the received data, the processed data, the masked or filtered data, or the processing units.

Abstract: A method, device, and computer program storage product for generating a query to extract clinical features into a set of electronic medical record (EMR) tables based on clinical knowledge. A knowledge tree is constructed according to a set of clinical knowledge data. An EMR graph corresponding to a set of EMR tables is obtained. The EMR graph comprises at set of table nodes and a set of attribute nodes. The set of table nodes and the set of attribute nodes represent a structure of each EMR table in the set of EMR tables and a reference relationship among attributes of set of EMR tables. A plurality of sub-queries is generated based on the knowledge tree and the EMR graph. At least one query is generated by combining the plurality of sub-queries according to the knowledge tree.

Abstract: According to an embodiment of the present disclosure for solving the aforementioned problem, disclosed is a computer program stored in a computer-readable storage medium executable by one or more processors, in which when the computer program is executed by one or more processors of a computer device, the computer program allows the one or more processors to perform the following operations for data processing, and the operations may include: an operation of generating a plurality of transformed data based on each of a plurality of data included in a data set; an operation of generating a test data set based on the plurality of data and the plurality of transformed data; and an operation of testing the performance of the model by calculating the test data set by using the model.

Abstract: An autonomous navigation system for a vehicle includes a controller configured to control the vehicle, sensors configured to detect objects in a path of the vehicle, nonvolatile memory including an artificial neural network configured to classify the objects detected by the sensors, and a processor. The artificial neural network includes a series of neurons in each of an input layer, at least one hidden layer, and an output layer. The memory includes instructions which, when executed by the processor, cause the processor to train the artificial neural network on a first task, identify, utilizing a contrastive excitation backpropagation algorithm, important neurons for the first task, identify, utilizing a learning algorithm, important synapses between the neurons for the first task based on the important neurons identified, and rigidify the important synapses to achieve selective plasticity of the series of neurons in the artificial neural network.

Abstract: Embodiments of the present disclosure are directed to a system, methods, and computer-readable media for facilitating stylistic expression transfers in machine translation of source sequence data. Using integrated loss functions for style transfer along with content preservation and/or cross entropy, source sequence data is processed by an autoencoder trained to reduce loss values across the loss functions at each time step encoded for the source sequence data. The target sequence data generated by the autoencoder therefore exhibits reduced loss values for the integrated loss functions at each time step, thereby improving content preservation and providing for stylistic expression transfer.

Abstract: A method for restoring a compressed image according to an embodiment of the present disclosure includes receiving monochrome image data and low resolution color image data generated from an original color image of the monochrome image data, decoding the monochrome image data and generating a low resolution monochrome image, decoding the low resolution color image data generating a low resolution color image; processing the low resolution monochrome image and generating a high resolution monochrome image in accordance with a super resolution imaging neural network; and generating a high resolution color image based on the low resolution color image and the high resolution monochrome image in accordance with a colorization imaging neural network. The imaging neural network of the present disclosure may be a deep neural network generated by machine learning, and images may be input and output in the Internet of Things environment using a 5G network.

Abstract: Embodiment of the invention are directed to transmitting signals between neurons of a hardware-implemented, spiking neural network (or SNN). The network includes neuronal connections, each including a synaptic unit connecting a pre-synaptic neuron to a post-synaptic neuron. Spikes received from the pre-synaptic neuron of said each neuronal connection are first modulated, in frequency, based on a synaptic weight stored on said each synaptic unit, to generate post-synaptic spikes, such that a first number of spikes received from the pre-synaptic neuron are translated into a second number of post-synaptic spikes. At least some of the spikes received from the pre-synaptic neuron may, each, be translated into a train of two or more post-synaptic spikes. The post-synaptic spikes generated are subsequently transmitted to the post-synaptic neuron of said each neuronal connection. The novel approach makes it possible to obtain a higher dynamic range in the synapse output.

Abstract: A processor-implemented data processing method includes: predicting whether there will be an inefficient section, of a neural network set to be implemented, during a processing of data, based on a hardware configuration for processing the data; adjusting a layer parameter corresponding to the inefficient section of the neural network; and processing the data using the neural network with the adjusted layer parameter.

Abstract: Embodiments of the present invention provide an artificial neural network system for feature pattern extraction and output labeling. The system comprises a first spiking neural network and a second spiking neural network. The first spiking neural network is configured to autonomously learn complex, temporally overlapping features arising in an input pattern stream. Competitive learning is implemented as spike timing dependent plasticity with lateral inhibition in the first spiking neural network. The second spiking neural network is connected with the first spiking neural network through dynamic synapses, and is trained to interpret and label the output data of the first spiking neural network. Additionally, the labeled output of the second spiking neural network is transmitted to a computing device, such as a central processing unit for post processing.

Abstract: The present disclosure relates to systems and methods for using transfer learning in log parsing neural networks. In one implementation, a system for training a neural network to parse unstructured data may include a processor and a non-transitory memory storing instructions that, when executed by the processor, cause the system to: receive unstructured data; apply a classifier to the unstructured data to determine that the unstructured data comprises a new category of unstructured data; in response to the determination, identify an existing category of unstructured data similar to the new category; based on the identified existing category, select a corresponding neural network; reset at least one weight and at least one activation function of the corresponding neural network while retaining structure of the corresponding neural network; train the reset neural network to parse the new category of unstructured data; and output the trained neural network.