Abstract: A method, system, and computer program product for learning entity resolution rules for determining whether entities are matching. The method may include receiving historical pairs of entities. The method may also include determining a set of rules for determining whether a pair of entities are matching, where the set of rules comprises a plurality of conditions. The method may also include developing, using a deep neural network, an entity resolution model based on the historical pairs of entities. The method may also include receiving a new pair of entities. The method may also include applying the entity resolution model to the new pair of entities. The method may also include determining whether one or more rules from the set of rules are satisfied for the new pair of entities. The method may also include categorizing the new pair of entities as matching or not matching.

Abstract: A multiply-accumulate calculation device includes: a plurality of first multiple calculation elements configured to generate first output signals by multiplying a first input signal corresponding to an input value by a weight and output the first output signals; and an accumulate calculation unit configured to calculate a sum of the first output signals output from the plurality of first multiple calculation elements in a calculation period from a point in time at which transition to a steady state has occurred after transient responses caused by charging to parasitic capacitors of the plurality of first multiple calculation elements according to input of the first input signal to a point in time after transient responses caused by discharging from the parasitic capacitors of the plurality of first multiple calculation elements according to input of the first input signal have started to be generated.

Abstract: Record clustering is performed by learning from verified clusters which are used as the source of training data in a deduplication workflow utilizing supervised machine learning.

Abstract: Described herein are systems and methods for efficiently processing large amounts of data when performing complex neural network operations, such as convolution and pooling operations. Given cascaded convolutional neural network layers, various embodiments allow for commencing processing of a downstream layer prior to completing processing of a current or previous network layer. In certain embodiments, this is accomplished by utilizing a handshaking mechanism or asynchronous logic to determine an active neural network layer in a neural network and using that active neural layer to process a subset of a set of input data of a first layer prior to processing all of the set of input data.

Abstract: The learning device 80 includes an input unit 81 and an imitation learning unit 82. The input unit 81 receives input of a type of a reward function. The imitation learning unit 82 learns a policy by imitation learning based on training data. The imitation learning unit 82 learns the reward function according to the type by the imitation learning, based on a form defined depending on the type.

Abstract: In an aspect, an apparatus for predicting downhole conditions is disclosed. The apparatus comprises at least a processor and a memory communicatively connected to the at least a processor. The memory containing instructions configuring the at least a processor to receive a condition datum. The memory additionally contains instructions configuring the at least a processor to produce a measured downhole condition using at least a sensor. The memory instructs the processor to convert the condition datum and the measured downhole conditions into a cleansed data format using a data conversion module. The processor is instructed by the memory to identify a flagged data as a function of the cleansed condition datum and the cleansed measured downhole conditions using a downhole machine learning model. The memory instructs the processor to generate a predicted downhole condition as a function of the flagged data.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating an output sequence from an input sequence. In one aspect, one of the systems includes an encoder neural network configured to receive the input sequence and generate encoded representations of the network inputs, the encoder neural network comprising a sequence of one or more encoder subnetworks, each encoder subnetwork configured to receive a respective encoder subnetwork input for each of the input positions and to generate a respective subnetwork output for each of the input positions, and each encoder subnetwork comprising: an encoder self-attention sub-layer that is configured to receive the subnetwork input for each of the input positions and, for each particular input position in the input order: apply an attention mechanism over the encoder subnetwork inputs using one or more queries derived from the encoder subnetwork input at the particular input position.

Abstract: In an aspect an apparatus for location monitoring. The apparatus includes at least a processor and a memory communicatively connected to the at least a processor. The memory instructs the processor to receive situational location data; receive a query as a function of the situational location data; generate a response as a function of the query; generate an optimal monitoring protocol as a function of the situational location data, wherein generating the optimal monitoring protocol includes training an optimal monitoring protocol machine learning model using optimal monitoring protocol training data, wherein the optimal monitoring protocol training data includes inputs correlated to outputs; and display the response using a display device.

Abstract: In at least one embodiment, differentiable neural architecture search and reinforcement learning are combined under one framework to discover network architectures with desired properties such as high accuracy, low latency, or both. In at least one embodiment, an objective function for search based on generalization error prevents the selection of architectures prone to overfitting.

Abstract: Embodiments of the present disclosure implement a stochastic neural network (SNN) where a subset of the nodes in the network are selectively activated based on sampling a plurality of computational paths traversing the network and based on different minimum thresholds for activation. In various embodiments, an output of the stochastic neural network is a sequence of the sampled plurality of computational paths with a corresponding sequence of output values that represent approximations of the output of the stochastic neural network. The nodes can include at least one input node, at least one output node and at least two hidden nodes, wherein the hidden nodes are positioned between the input node and the output node, and wherein sampling the plurality of computational paths involves initiating each of the plurality of computational paths from a first of the hidden nodes, wherein the first of the hidden nodes has been activated by a previous computational path.

Abstract: A computer-implemented bidding method, computer equipment and a storage medium are provided. The computer-implemented bidding method includes: training a CatBoost regression model through a historical bidding data set, where the historical bidding data set includes bidding configuration parameters as an input of the model and a difference between a first quote and a final quote as an output of the model; and inputting current basic bidding parameters into the trained CatBoost regression model, and outputting values of optimized bidding configuration parameters to configure bidding rules for bidding participants. The computer-implemented bidding method can help a purchaser to purchase a required product at a relatively low price, thereby saving the purchase cost.

Abstract: A method of designing a nanostructure, comprises: receiving a far field optical response and material properties; feeding the synthetic far field optical response and material properties to an artificial neural network having at least three hidden layers; and extracting from the artificial neural network a shape of a nanostructure corresponding to the far field optical response.

Abstract: An apparatus for data ingestion and manipulation, the apparatus including at least a processor and a memory communicatively connected to the at least a processor, the memory containing instructions configuring the at least a processor to receive a resource data file from one or more data acquisition systems, classify the resource data file to one or more educational categorizations, generate an educational module as a function of the resource data file and the classification of the educational categorizations wherein the education module comprises one or more machine learning models, retrieve a user profile of a plurality of user profiles as a function of a user input, create user-specific outputs as a function of the educational module, the user profile, and a conversational input and generate a virtual avatar model as a function of the user specific outputs.

Abstract: Apparatus, methods, and systems for cross-domain time series data conversion are disclosed. In an example embodiment, a first time series of a first type of data is received and stored. The first time series of the first type of data is encoded as a first distributed representation for the first type of data. The first distributed representation is converted to a second distributed representation for a second type of data which is different from the first type of data. The second distributed representation for the second type of data is decoded as a second time series of the second type of data.

Abstract: Example implementations described herein involve systems and methods to substantially simultaneously orchestrate machine learning models over multiple resource constrained control edge devices, so that the overall system is more agile to changes in events and environmental conditions where the models have been deployed. The example implementations described herein involve multiple processes that when executed, determine a list of edge devices to be updated along with the corresponding models based on correlation.

Abstract: The present subject matter refers a method for developing machine-learning (ML) based tool. The method comprises initializing an input dataset for undergoing ML based processing. The input dataset is pre-processed by a first model to harmonize features across the dataset. Thereafter, the dataset is annotated by a second model to define a labelled data set. A plurality of features are extracted with respect to the data set through a feature extractor. A selection of at-least a machine-learning classifier is received through an ML training module to operate upon the extracted features and classify the dataset with respect to one or more labels.

Abstract: Disclosed is a method of training an object prediction model by using input data and a discrimination label including a plurality of discrimination information by a computing device including at least one processor which is a training method including: generating a prediction label based on the input data by using the prediction model; generating a loss value based on a discrimination label corresponding to the input data and the prediction label; and training the prediction model based on the loss value.

Abstract: The disclosed embodiments concern methods, apparatus, systems and computer program products for determining and displaying pedigrees based on IBD data. Some implementations use a probabilistic relationship model to obtain various likelihoods of various potential relationships based on pairwise IBD data and pairwise age data. Some implementations build large pedigrees by combining smaller pedigrees. Some implementations display pedigree graphs with various features that are informative and easy to understand.

Abstract: Described is an attack system for generating perturbations of input signals in a recurrent neural network (RNN) based target system using a deep reinforcement learning agent to generate the perturbations. The attack system trains a reinforcement learning agent to determine a magnitude of a perturbation with which to attack the RNN based target system. A perturbed input sensor signal having the determined magnitude is generated and presented to the RNN based target system such that the RNN based target system produces an altered output in response to the perturbed input sensor signal. The system identifies a failure mode of the RNN based target system using the altered output.

Abstract: Disclosed herein is a visual analytics system and method that provides a proactive and predictive environment in order to assist decision makers in making effective resource allocation and deployment decisions. The challenges involved with such predictive analytics processes include end-users' understanding, and the application of the underlying statistical algorithms at the right spatiotemporal granularity levels so that good prediction estimates can be established. In the disclosed approach, a suite of natural scale templates and methods are provided allowing users to focus and drill down to appropriate geospatial and temporal resolution levels. The disclosed forecasting technique is based on the Seasonal Trend decomposition based on Loess (STL) method applied in a spatiotemporal visual analytics context to provide analysts with predicted levels of future activity.