Abstract: Systems and computer-implemented methods facilitating automatic report generation, population and information prompting employing artificial intelligence technology are provided.

Abstract: Methods and apparatus disclosed herein autonomously evaluate replacement algorithms. An example system associated with a live environment executing a current algorithm includes at least one memory, instructions, and processor circuitry to execute the instructions to manage execution of the current algorithm in the live environment, the live environment associated with a deployment platform, and manage, based on receipt of at least one potential replacement algorithm, a shadow environment causes execution of the at least one potential replacement algorithm in the shadow environment, the shadow environment instantiated separately from the live environment, the shadow environment utilizes data from the live environment during execution of the at least one potential replacement algorithm in the shadow environment.

Abstract: Provided is a system and method for generating an SQL operation from a natural language input using neural networks which works on data from a database that has not been seen before by the neural networks. In one example, the method may include receiving a natural language input, determining, via a first predictive module, a data structure of a SQL operation based on text included in the natural language input, filling in, via a second predictive module, one or more empty slots of the data structure of the SQL operation with text based on column names of a database and the text included in the natural language input, and storing the filled-in data structure of the SQL operation in a storage.

Abstract: A method of operating a robotic system includes: receiving first image data representative of a package surface; identifying a pair of edges based on the first image data; determining a minimum viable region based on the pair of edges; receiving second image data representative of the package after the lift; and creating registration data based on the first and second image data.

Abstract: Aspects of the disclosure relate to identifying and utilizing head and eye gestures, using a wearable computing device. The wearable computing device may include an electrooculography electrode configured to provide first feedback to one or more processors as well as an inertial measurement unit configured to provide second feedback to the one or more processors. The first feedback may correspond to eye movement, and the second feedback may correspond to head movement. The wearable computing device may also include the one or more processors which may be configured to use the first feedback and these second feedback to determine whether the head movement and eye movement correspond to a trigger for an action as well as to cause the wearable computing device to perform the action based on the determination.

Abstract: The present application relates to the technical field of computers, and provides a user request response method, a user request response apparatus, a computer-readable storage medium, and an electronic device. The user request response method comprises: upon the reception of a user request sent from a first-type terminal, sending reminder messages to a plurality of second-type terminals according to the user request; receiving message receipts for the reminder messages within a preset time period; selecting a target message receipt from among the message receipts according to a preset rule; and using the second-type terminal corresponding to the target message receipt as a target second-type terminal and determining to respond to the user request according to the target second-type terminal.

Abstract: Data about clients, items included in an item inventory, executable code for recommendation processes, and executable code for feature selection process(es), are stored. The recommendation processes are used to identify items in the item inventory that are recommended for specific clients, and the feature selection process(es) is/are used to select features that are displayed to entities tasked with selecting items from the item inventory for clients. Selection feedback and client feedback are received. One or more feature selection process(es) is/are modified based on the selection feedback and/or the client feedback, to thereby change which, where or/or how certain portion of the data about clients is displayed to the entities tasked with selecting items from the item inventory for the clients, and/or change which, where and/or how certain portion of the data about items is displayed to the entities tasked with selecting items from the item inventory for the clients.

Abstract: Techniques for making machine learning inference calls for database query processing are described. In some embodiments, a method of making machine learning inference calls for database query processing may include generating a first batch of machine learning requests based at least on a query to be performed on data stored in a database service, wherein the query identifies a machine learning service, sending the first batch of machine learning requests to an input buffer of an asynchronous request handler, the asynchronous request handler to generate a second batch of machine learning requests based on the first batch of machine learning requests, and obtaining a plurality of machine learning responses from an output buffer of the asynchronous request handler, the machine learning responses generated by the machine learning service using a machine learning model in response to receiving the second batch of machine learning requests.

Abstract: A method for initiating and automatically improving model-driven operations in a low-data scenario includes creating a regression model using pre-operation data prior to initiating the model-driven operations, using the regression model to initiate and perform the model-driven operations during an operational stage, collecting operational data during the operational stage, creating a first artificial neural network model using the operational data, transitioning from using the regression model to perform the model-driven operations to using the first artificial neural network model to perform the model-driven operations responsive to the operational data satisfying a first sufficiency threshold.

Abstract: Provide an information processing apparatus including a determining unit configured to determine whether to generate a learned model that estimates a value relating to a component included in a device, based on at least one of a measurement value measured by each sensor of the device, an anomaly diagnosis result of the device, a result of controlling the device, or a condition relating to the device.

Abstract: Methods and systems for monitoring a plurality of components of a pump in an industrial environment include a data acquisition circuit structured to interpret a plurality of detection values, each of the plurality of detection values corresponding to at least one of a plurality of input sensors operationally coupled to the pump and communicatively coupled to the data acquisition circuit; a data processing circuit structured to utilize at least one of the plurality of detection values to perform at least one noise processing operation on at least a portion of the plurality of detection values; a signal evaluation circuit structured to determine a pump performance parameter in response to the noise processed plurality portion of the plurality of detection values; and a response circuit structured to perform at least one operation in response to the pump performance parameter.

Abstract: Embodiments include technologies to apply at least one machine learning model to features of a search query, features of a searcher user, features of a searchee content item, and features of a searchee user, produce a first outcome prediction that represents a probability of a first objective relating to engagement of the searcher user with a content item in an online system and a second outcome prediction that represents a probability of a second objective relating to engagement of the searchee user with the online system responsive to the engagement of the searcher user with the content item, apply a multi-objective optimization solver to the first objective, the second objective and an outcome prediction that is a combination of the first outcome prediction and the second outcome prediction, and generate a serving function for a search engine based on the first objective, the second objective, and the outcome prediction.

Abstract: An advice presentation system that combines functions of a terminal such as a smartphone, advice data such as famous or wise sayings that have been stored in a database, and functions such as a communications channel together, and provides appropriate advice according to the emotions of a user at that time. Firstly, emotion-related data pertaining to the emotions of the user is entered from the terminal. The advice presentation system connects to a web server via Internet connection, and emotion data corresponding to the entered emotion-related data is selected and extracted from an emotion database by using an emotion data selection means for an application server. At least one piece of advice data stored in an advice database is selected and extracted by an advice data selection means for the application server, on the basis of the selected emotion data. The selected advice data is sent to the terminal of the user, and famous or wise sayings, etc., are presented from an advice display means as advice data.

Abstract: Proactively detecting one or more handover error in a communication link between base stations [14y, 14z] in a wireless communication network. A method may include determining at least one possible solution to the one or more handover error using an artificial intelligence engine 76. A method may include implementing the at least one possible solution to correct the one or more handover error. The artificial intelligence engine 76 may include an artificial intelligence algorithm, a machine learning algorithm, a deep learning algorithm, a neural network algorithm; and/or big data analysis algorithm. The artificial intelligence engine 76 may utilize at least one database [92, 94, 96] including a plurality of solutions. The plurality of solutions may include solutions to errors in communication networks that occurred in the past. The artificial intelligence engine 76 may successively implement each of the possible solutions until the one ore more handover error is corrected.

Abstract: Data encryption and Human Pose Estimation based on imaging a body segment. A key for encrypting a data file is generated based on image data that represent a unique biometric feature of a body segment of a user or motion of the user. An image engine executes artificial intelligence to identify matching image data for decrypting the data file. The image engine is further trained to predict changes in image data due to aging, stress, and the like. An avatar associated with the user, which is generated based on a movement pattern of the user, is configurable for generating an encryption key and for use in an avatar-based language.

Abstract: The improved active noise cancellation system for forced air heating or cooling systems onboard aircraft employs a duct having a proximal end coupled to the fan unit to entrain the airflow stream in the direction of a distal end of the duct. A reference sensor is positioned within the proximal end of the duct. A means is provided for injecting an audio frequency control signal into the airflow stream in a manner that does not substantially impede the airflow stream. An error sensor is positioned at the distal end of the duct where it is responsive to sounds carried by the airflow stream, including the audio frequency control signal. An electronic circuit coupled to the reference sensor and to the error sensor supplies a noise abating control signal to energize the control transducer and thereby substantially reduce at least one noise harmonic of the fan unit through destructive interference.

Abstract: A condition of an asset including one or more subsystems connected in series is monitored. Each subsystem includes one or more components connected in parallel. The asset has one or more jobs. A probability of the asset surviving a predetermined amount of time is determined based on the monitoring and one or more shared resources. The one or more shared resources are configured to be shared between the subsystems. A model is established using a threshold based heuristic maintenance policy. The model is configured to maximize a number of successful jobs that the asset is able to complete based on the determined probability. The one or more shared resources are allocated to the one or more subsystems based on the model.

Abstract: The invention provides a system and method for providing ttx-based categorization services and a categorized commonplace of shared information. Currency of the contents is improved by a process called conjuring/concretizing wherein users' thoughts are rapidly infused into the Map. As a new idea is sought, a goal is created for a search. After the goal idea is found, a ttx is concretized and categorized. The needs met by such a Map are prior art searching, competitive environmental scanning, competitive analysis study repository management and reuse, innovation gap analysis indication, novelty checking, technology value prediction, investment area indication and planning, and product technology comparison and feature planning.

Abstract: Methods and apparatuses for accurately determining a model, which is to be the basis of transfer learning, among a plurality of source models, are provided. According to an embodiment, an apparatus for determining a base model to be used for transfer learning to a target domain is provided. The apparatus comprises a memory which comprises one or more instructions and a processor which executes the instructions to construct a neural network model for measuring suitability of a plurality of pre-trained source models, measure the suitability of each of the source models by inputting data of the target domain to the neural network model, and determine the base model to be used for the transfer learning among the source models based on the suitability.

Abstract: Systems and methods are provided for generating new insights for multi-dimensional data. A plurality of insight definitions that define insights within multi-dimensional data can be received, where the multi-dimensional data include dimensions organized in a hierarchical structure, and the insight definitions include members across the hierarchical dimensions that define a slice of the multi-dimensional data. Descriptive data about a candidate insight defined by at least one of the insight definitions that corresponds to a candidate slice of the multi-dimensional data can be received. A new insight definition based on the candidate insight can be generated, where the new insight definition corresponds to a new slice of the multi-dimensional data, and the new slice of the multi-dimensional data is proximate to the candidate slice of the multi-dimensional data with respect to the hierarchical structure.

Abstract: Methods, systems, and computer-readable media for generating access control policies using static analysis are disclosed. An access control policy generator performs static analysis of program code of a software product. The static analysis identifies one or more calls to one or more external components in the program code. The access control policy generator determines a mapping of the one or more calls to one or more actions. The one or more actions are selected from a plurality of known actions supported by an access control policy manager. The access control policy generator generates an access control policy associated with the software product. The access control policy comprises one or more permissions with respect to the one or more external components. The access control policy permits the software product to access the plurality of external components using the access control policy manager during execution of the software product.

Abstract: A dynamic support system includes a control system for controlling inflation and deflation of at least one actuator having an inlet connectable to the a control unit of the dynamic support system. The control unit may be in communication with a sensor and may control inflation and deflation of the at least one actuator in response to information provided by the sensor.

Abstract: A trained machine learning model(s) is used to determine scores indicative of probabilities that certain types of user input will be provided to a player's game controller while playing a video game in order to compensate for latency between player action and player perception of video game content relating to the player action. In an example process, sensor data received from a client machine and/or game state data received from a video game is provided as input to a trained machine learning model(s), and a score as output therefrom, the score relating to a probability that a type of user input will be provided to a player's game controller. In this manner, game control data corresponding to the type of user input can be generated based on the score and provided to the video game as input before actual game control data is even received, thereby compensating for latency.

Abstract: The present disclosure relates to systems and methods customizing electronic communications. A future event associated with a first user may be determined, and a second user that is associated with the first user may be identified. A plurality of communications involving the first user and the second user may be analyzed. A selection rule may be applied based on the analyzed plurality of communications, the selection rule identifying content from the database. Content from the database may be selected based on the application of the selection rule. An electronic message may be provided to the first user identifying the future event, and the selected content may be provided to the first user.

Abstract: A System and Method is provided that enable identifying cyber security attacks using observation and monitoring of end point activity. By following and monitoring the wireless connection related activities of endpoint devices as they cycle through various steps leading to establishing a connection to the secure network, a knowledge base is established in the cloud by analysis of the actions, and communication to build the confidence that the users of the network are where they should be. In one embodiment, no access is provided until a user presents valid credentials. Based on these credentials the network then builds a specific path based on access controls, tunnels or other techniques to control the user’s communication and access to specific targets within the network.

Abstract: A method is described herein, comprising registering an event at a first processing unit of a processing facility comprising a plurality of processing units, using a coincidence probability array and an event probability to identify a second processing unit of the plurality of processing units based on the event, determining whether the second processing unit experienced a coincident event, if the second processing unit experienced a coincident event, remediating a condition of the second processing unit that caused the coincident event, and updating the coincidence probability array based on the event.

Abstract: An optimization method, apparatus, computer device and storage medium for engine model. The method includes: setting an initial population according to a predetermined number of solutions; obtaining a first output set of the initial population on the engine model; determining a first fitness value; operating selection, crossover, and mutation on the initial population according to the first fitness value to obtain an evolved population, and the initial population and the evolved population constitute a to-be-selected population; obtaining a second output set of the evolved population on the engine model; determining a second fitness value; selecting engine model parameters in the to-be-selected population to form a new population according to the second fitness value; and optimizing the engine model by using the engine model parameters in the new population if a termination condition is met.

Abstract: A computer-implemented position-centric personnel modeling apparatus, system, and method may be provided. A nodal hierarchy may be created to correlate with the personnel structure of a business enterprise, wherein each node may be a cyberspace representation of an individual within the personnel structure of the business enterprise. Attributes correlating to individual personnel in the business enterprise may be mapped to a respective node. Objectives may subsequently be tasked within the model and the nodal hierarchy may be rearranged based on the results. Results from the model may optionally be reflected by a business enterprise.

Abstract: A technique for verifying data structure consistency across computing environments includes computing a first signature for a data structure of an application subject to checkpointing corresponding to a first computing environment residing on a first computer. A second signature for the data structure of the application corresponding to a second computing environment residing on a second computer is computed. The first and second signatures are compared to determine whether a change to the data structure exists. Responsive to a lack of change to the data structure based on the comparison, a mobility operation is enabled for the application between the first computer and the second computer.

Abstract: According to some embodiments, an available algorithm data store may contain information about a pool of available algorithms. An algorithm selection platform coupled to the available algorithm data store may access the information about the pool of available algorithms and compare the information about each of the pool of available algorithms with at least one requirement associated with the current algorithm executing in the real environment. The algorithm selection platform may then automatically determine algorithm execution context information and, based on said comparison and the algorithm execution context information, select at least one of the pool of available algorithms as a potential replacement algorithm. An indication of the selected at least one potential replacement algorithm may then be transmitted (e.g., to be evaluated in a shadow environment by an algorithm evaluation platform).

Abstract: An end-to-end software framework is provided for information extraction, enrichment, and caching in applications that involve exploring real-world data through a camera-based viewport in an augmented reality (AR) setting. It allows users to gather supplementary information through cross-modal data fusion against cloud-based datastores. Additional context is provided by retrieving data units from remote datastores that are relevant to the queried item. Given that such datastores often do not fit in main memory, a client-server architecture is used. Users interact with real-world objects via a camera-enabled client interface, and queries are executed using gesture-based interactions as supplementary information is retrieved from a back-end server (if necessary) and overlaid in the client's live camera-view.

Abstract: A system for retrieving information from an instructional document, includes a processor configured to: receive a query from a user; compare the query with one or more text sections in the instructional document; obtain, from the one or more text sections, top x text sections relevant to the query using a pre-trained encoder; compare the query with one or more images in the instructional document; obtain, from the images, top y images relevant to the query; generate top y image-text sections based on the top y images; obtain top k sections from the top x text sections and the top y image-text sections; obtain one or more most relevant sections from the top k sections using a domain-specific pre-trained encoder and a sequential classification model; and generate an answer based on the one or more most relevant sections and device context information using the domain-specific pre-trained encoder.

Abstract: Systems, methods, and related technologies for classification are described. Network traffic transmitted by a first device is obtained. A set of features is determined based on the network traffic. A first classification for the device is determine a first classification for the first device based on the set of features. The first classification is associated with a first classification level. A second machine learning model is identified based on the first classification. The second machine learning model is associated with the first classification. A second classification for the first device is determined based on the second machine learning model. The second classification is associated with a second classification level. At least one of the first classification and the second classification is stored.

Abstract: Methods and systems for fault identification and mitigation in an engine system. A state observer obtains current state information from the engine system, and a feature calculator uses data obtained from the state observer to calculate one or more feature indicators, which are monitored by a health estimator for the occurrence of a change using one or more change probability models. When the health estimator identifies a change, a fault isolator determines a component of the engine system that is subject to fault or health deterioration.

Abstract: A system maintains a knowledge layout to support the building of event and analytics models in parity. The system uses the event models to provide a snapshot of the relevant conditions present when a challenge event occurs. The system uses the analytics models to select one or more actions (which may include robotic tasks) to respond to the challenge condition. In some cases, the system may render continued response compulsory until a successful response to the challenge event is achieved.

Abstract: An approach for tuning a data management system is provided. A processor retrieves one or more samples pairs of record data. A processor determines an error rate regarding the performance of a matching operation of a data management system based on the one or more sample pairs. In response to the error rate being above a value, a processor adjusts at least one threshold value associated with the matching operation of the data management system.

Abstract: Illustrative storage manager and media agent are enhanced to interoperate with deduplication appliances. Advantages are realized when making secondary and tertiary copies and also when restoring from a deduplication appliance. Tiered indexing minimizes how much data is retained and stored at media agents. Tiered indexing enables media agents to efficiently extract needed information from deduplication appliances to make tertiary copies and to restore backed up copies. Interoperability techniques include media agents generating separate data streams to the deduplication appliance. Each data stream carries a different kind of data, e.g., payload data, metadata content, or high-level index information. On initial backup, the media agent instructs the deduplication appliance to deduplicate the payload data stream but not the other data streams, thus intelligently applying resources to data most likely to benefit from deduplication.

Abstract: A network interface card (NIC) and a method for stablishing a connection between virtual machines of a network. The NIC includes: a programmable switching ASIC (application-specific integrated circuit), a central processing unit (CPU), multiple Ethernet controllers, and multiple on-board transceivers functioning as external ports. The switching ASIC functions as a switch that manipulates data traffic within the NIC including by switching the data traffic between and among the CPU, the Ethernet controllers, and the on-board transceivers. The method includes: installing rules that route a Synchronize (SYN) packet from a source virtual machine (VM) through a software engine, appending a signed cookie to the SYN packet; verifying that a policy represented by the signed cookie appended to the SYN packet matches a policy of a destination VM; and returning the SYN packet to the source VM which establishes a connection between the source VM and the destination VM.

Abstract: A method, system, and medium are provided for supplying contextual help to a user of an application. A problem encountered by the user while using the application is identified in real time. Identifying the problem may include identifying the current user task, based at least on application checkpoints encountered by the user. In response to identifying the problem encountered by the user in real time, the user is provided with short-term help and long-term help that address the problem. The short-term and long-term help may be identified based at least in part on mappings that associate user tasks with help information applicable to the user tasks.

Abstract: For machine learning data reduction and model optimization,a method randomly assigns each data feature of a training data set to a plurality of solution groups. Each solution group has no more than a solution group number k of data features and each data feature is assigned to a plurality of solution groups. The method identifies each solution group as a high-quality solution group or a low-quality solution group. The method further calculates data feature scores for each data feature comprising a high bin number and a low bin number. The method determines level data for each data feature from the data feature scores using a fuzzy inference system. The method identifies an optimized data feature set based on the level data. The method further trains a production model using only the optimized data feature set. The method predicts a result using the production model.

Abstract: Medical devices and related systems and methods are provided. A method of estimating a physiological condition involves determining a translation model based at least in part on relationships between first measurement data corresponding to instances of a first sensing arrangement and second measurement data corresponding to instances of a second sensing arrangement, obtaining third measurement data associated with the second sensing arrangement, determining simulated measurement data for the first sensing arrangement by applying the translation model to the third measurement data, and determining an estimation model for a physiological condition using the simulated measurement data, wherein the estimation model is applied to subsequent measurement output provided by an instance of the first sensing arrangement to obtain an estimated value for the physiological condition.

Abstract: An apparatus and method are described herein for training a similarity model that is used to predict similarity between item pairs. In one example, the apparatus is configured to train (build) the similarity model by using machine learning (e.g., a multivariate multiple liner regression process) that utilizes an independent variable including metadata (e.g., title, genre, writer, plot keywords) associated with training items, and two dependent variables including user contributed similarity scores for training item pairs, and collaborative filtering similarity scores for the training item pairs. Then, the apparatus uses the trained model to predict similarity between items.

Abstract: Systems and methods are described for tuning parameter values of a processing pipeline in a streaming data processing system. In order to determine an optimal set of parameter values for a particular processing pipeline, a processing pipeline can be implemented with different sets of parameter values. A performance metric can be measured for each implementation to measure the performance of the processing pipeline with regards to a particular set of parameter values. The performance metrics for each implementation can be compared in order to determine optimal performance metrics. The processing pipeline can be implemented based on an optimal set of parameter values that correspond to the optimal performance metrics.

Abstract: Computational systems and methods employ characteristics of a quantum processor determined or sampled between a start and an end of an annealing evolution per an annealing schedule. The annealing evolution can be reinitialized, reversed or continued after determination. The annealing evolution can be interrupted. The annealing evolution can be ramped immediately prior to or as part of determining the characteristics. The annealing evolution can be paused or not paused immediately prior to ramping. A second representation of a problem can be generated based at least in part on the determined characteristics from an annealing evolution performed on a first representation of the problem. The determined characteristics can be autonomously compared to an expected behavior, and alerts optionally provided and/or the annealing evolution optionally terminated based on the comparison. Iterations of annealing evolutions may be performed until an exit condition occurs.

Abstract: This application provides a data stream identification method and apparatus and belongs to the field of Internet technologies. The method includes: obtaining packet transmission attribute information of N consecutive packets in a target data stream; generating feature images of the packet transmission attribute information of the N consecutive packets based on the packet transmission attribute information of the N consecutive packets; and inputting the feature images into a pre-trained image classification model, to obtain a target application identifier corresponding to the target data stream. According to this application, accuracy of identifying an application identifier corresponding to a data stream can be improved.

Abstract: A system that controls a combustion engine stores network vectors in a memory that represent diverse and distinct spiking neural networks. The system decodes the network vectors and trains and evaluates the spiking neural networks. The system duplicates selected network vectors and crosses-over the duplicated network vectors that represent modified spiking neural networks. The system mutates the crossed-over duplicated network vectors by randomly modifying one or more portions of the crossing-over duplicated network vectors. The system meter exhaust gas into an intake manifold when an engine temperature exceeds a threshold, an engine load exceeds a threshold, an engine's rotation-per-minute rate exceeds a threshold, and a fuel flow exceeds a threshold. The system modifies fuel flow into an engine's combustion chamber on a cycle-to-cycle basis by the trained spiking neural network.

Abstract: In various example embodiments, a system and methods are presented for converting query structures for information retrieval from graph-based data structures. The systems and methods receive a natural language query including a set of terms and generate an intermediate semantic relationship of the set of terms of the natural language query. The systems and methods generate a graph query including graph terms corresponding to the set of terms of the natural language query defined by a graph database. The systems and methods search one or more datasets associated with the graph database using the graph query and return a set of results based on the graph query.

Abstract: Disclosed is a method for validating a hardware system by a model thereof, which method comprises: providing reference structures and determining, in the model, sub-structures, each of which is structurally equivalent to one of the reference structures; extracting, from the model, input cones for each sub-structure; creating monopartite candidate graphs by mapping the bipartite sub-structure and the respective input cones to one of the candidate graphs; creating, for each candidate graph, a match vector, each dimension of the match vector comprising a count of occurrences, in the candidate graph, of a different one of predetermined graphlets; clustering, on the basis of similarity of the match vectors, the candidate graphs in clusters; and selecting, from each of the clusters, one candidate graph and determining a functional behaviour of the respective sub-structure of the selected candidate graph for validating the hardware system.

Abstract: The present teaching relates to method, system, medium, and implementations for identifying k nearest neighbors. One or more KNN lists corresponding to one or more source data points are received. Each KNN list includes K neighbors of a source data point and each of the K neighbors is a data point represented by an index. Neighbor pairs and reverse neighbor pairs are generated based on the one or more KNN lists. The neighbor pairs and reverse neighbor pairs having the same source data point are grouped to generate a grouped pairs of neighbors for the source data point. A local join operation is performed based on grouped pairs of neighbors for each source data point to generate a combined neighborhood for the source data point, which is then sent to a KNN server, where combined neighborhoods generated by multiple local join executors are integrated to update a plurality of global KNN lists.

Abstract: Residual semi-recurrent neural networks (RSNN) can be configured to receive both time invariant input and time variant input data to generate one or more time series predictions. The time invariant input can be processed by a multilayer perceptron of the RSNN. The output of the multilayer perceptron can be used as an initial state for a recurrent neural network unit of the RSNN. The recurrent neural network unit can also receive time invariant input, and process the time invariant input with the time invariant input to generate an output. The outputs of the multilayer perceptron and the recurrent neural network unit can be combined to generate the one or more time series predictions.