Abstract: A method is provided for operating a multi-channel messaging system. The method may provide automated conversation across multiple communication channels associated with a user by: selecting a first communication channel from the multiple communication channels, wherein the first communication channel is hosted by a first communication server; receiving, via an existing user interface of the first communication channel, a first user input via as part of a conversation with a chatbot, wherein the chatbot comprises a communication data structure comprising a plurality of communication paths; selecting a communication path based on the first user input; and generating a first feedback in response to the first user input according to the selected communication path.

Abstract: Based on one metadata template determined from among a plurality of metadata templates, a screen including a scan image, a plurality of keys defined in the one determined metadata template, and value input fields each corresponding to a different one of the plurality of keys is displayed, and in a case where a user selects any one of the plurality of keys displayed in the screen, character string areas determined to match a data format of a value corresponding to the key are displayed in such a manner that the character string areas are recognizable as value candidates on the scan image.

Abstract: Real-time data from one or more sensors can be received about one or more active wellbore operations. A blended data set can be generated that includes the real-time data combined with historical data about previously completed wellbore drilling operations, and associated with performance attributes. Additionally, identification of a criteria of focus, types of parameters for optimization, and operation groupings for the blended data set can be received. Further, the criteria of focus, types of parameters for optimization and operation groupings for the data can be applied to the blended data set to determine an adjustment for an active wellbore drilling operation of interest. The adjustment for the active wellbore drilling operation of interest can be outputted.

Abstract: Telematics and external data relating to real-time driving of a population of drivers may be collected and used to calculate a driving pattern map. The driving pattern map is used to determine a driving quotient for individual drivers wherein the driving quotient is a relative score. The driving quotient may be displayed to the driver.

Abstract: Implementations of the present disclosure relate to visual programming for deep learning. A computer-implemented method comprises presenting a visual representation of an artificial neural network, the visual representation comprising graphical elements representing layers of the artificial neural network; in response to receiving a drag-and-drop operation on the graphical elements, modifying an intermediate representation of the artificial neural network, wherein the intermediate representation is independent of a deep learning framework and the drag-and-drop operation is configured to modify connections between the graphical elements; and modifying, based on the intermediate representation of the artificial neural network, code of the artificial neural network for a target deep learning framework.

Abstract: Embodiments in the present disclosure relate to computer network architectures and methods for predictive analysis using lookup tables as prediction models. The predictive analysis, including the generation of the lookup tables, performed by a predictive system of the present disclosure is driven entirely by a query language, such as Structured Query Language, in various embodiments. The predictive analysis, including the generation of the lookup tables, is performed without machine learning or generative artificial intelligence, in various embodiments.

Abstract: Provided are a method and an apparatus for testing AI chip computing performance, and a non-transitory computer-readable storage medium. The method includes: forming computing performance result data of a to-be-tested AI chip according to a plurality of items of simulation data formed in a development process of the to-be-tested AI chip; acquiring a function instruction set matched with a to-be-tested service function, wherein the function instruction set is composed of a plurality of instructions in a standard instruction set matched with the to-be-tested AI chip; and predicting computing time required by the to-be-tested AI chip to execute the to-be-tested service function according to the function instruction set and the computing performance result data.

Abstract: An apparatus and method for entity data aggregation and analysis is disclosed. The apparatus includes at least a processor and a memory communicatively connected to the at least a processor, wherein the memory contains instructions configuring the at least a processor to receive entity data from a data collection module, wherein the data collection module includes at least a sensor, determine an obstacle datum as a function of the entity data, classify the entity data into one or more obstacle groups as a function of the obstacle datum of the entity data, determine at least an obstacle resolution object as a function of an obstacle resolution template and generate the obstacle resolution data structure as a function of a template form field of an entity specific data structure template, the at least an obstacle resolution object and the one or more obstacle groups.

Abstract: Systems and methods of the present disclosure may receive, from a user computing device, a user-provided data record query including a natural language request for information associated with one or more data sources. User persona attributes of the user may be determined, such as a user role or security parameters or both. Based on the user persona attributes a context query may be generated to obtain context attributes associated with the user-provided query. The natural language request and the context attributes are input into the model orchestration large language model (LLM) to output instructions to machine learning (ML) agents based on the context attributes. The ML agents output responses associated with the user-provided data record query based on the instructions, and the responses are input into the model orchestration LLM to output to the user computing device a natural language response based on the context attributes.

Abstract: Embodiments are directed towards real time display of event records and extracted values based on at least one extraction rule, such as a regular expression. A user interface may be employed to enable a user to have an extraction rule automatically generate and/or to manually enter an extraction rule. The user may be enabled to manually edit a previously provided extraction rule, which may result in real time display of updated extracted values. The extraction rule may be utilized to extract values from each of a plurality of records, including event records of unstructured machine data. Statistics may be determined for each unique extracted value, and may be displayed to the user in real time. The user interface may also enable the user to select at least one unique extracted value to display those event records that include an extracted value that matches the selected value.

Abstract: A system for use in a connected kitchen allows a user to find and use traditional recipes as would be found in a recipe book and to have them automatically customized to their language, locality, dietary preferences and the appliances in their kitchen.

Abstract: Some systems, devices and methods detailed herein provide a system for use in determining metrics of a subject. The system can provide, as an output, a function-metric value determined based on a defined relationship between physiological measures and a chronological age.

Abstract: Systems and methods for presenting configurable machine learning systems through graphical user interfaces are disclosed. In an embodiment, a machine learning server computer stores one or more machine learning configuration files. A particular machine learning configuration file of the one or more machine learning configuration files comprises instructions for configuring a machine learning system of a particular machine learning type with one or more first machine learning parameters. The machine learning server computer displays through a graphical user interface, a plurality of selectable parameter options, each of which defining a value for a machine learning parameter. The machine learning server computer receives a particular input dataset.

Abstract: In a method for AI model optimization, an optimization device receives an original AI model and search configuration information that comprises a plurality of search items each indicating its search categories for performing optimization information search on the original AI model. The device obtains a plurality of search operators corresponding to the plurality of search items, and arranges the search operators in an operation sequence based on the search configuration information. The device then executes the search operators in the arranged operation sequence on the original AI model to obtain an optimized AI model. In the execution of the operation sequence, each search operator, except for the first search operator in the operation sequence, is executed utilizing operation results of a preceding search operator in the operation sequence, the operation results including generated network structures and search space information.

Abstract: A cross domain generalization system for industrial artificial intelligence (AI) applications is disclosed. A target encoder subsystem obtains target data from a target machine product and generates lower dimensional data for obtained target data using a target artificial intelligence (AI) model. The generated lower dimensional data are corresponding to a plurality of target embeddings data. The target encoder subsystem further applies the plurality of target embeddings data into a source classifier AI model. A source classifier subsystem predicts a quality of the target machine product by generating class labels for each of the plurality of target embeddings data based on a result of the classifier AI model. The goal of the present invention is to learn features or representations such that the correlation with a label space is similar both in source and target domains while being invariant of data distributions.

Abstract: An information processing apparatus includes a processor configured to acquire biological information of plural users determined to be within a predetermined area and operate an apparatus based on the biological information of the plural users.

Abstract: Systems and methods can be provided for predicting responses during communication sessions with network devices. In some implementations, systems and methods can facilitate predicting responses using machine learning techniques. Messages received through a platform can be stored in a repository. A machine learning model may be trained using the stored messages. When a terminal device is communicating with a network device in a communication session, the messages exchanged in the communication session and the machine learning model can be used to predict future responses in real-time. The predicted future responses can be presented at the terminal device. A predicted response can be selected at the terminal device. Upon selection, the selected predicted response is transmitted to the network device during the communication session.

Abstract: A non-transitory machine-readable medium can comprise executable instructions that, when executed by a processor, facilitate performance of operations, comprising: determining a workflow comprising a group of files of a distributed file system, determining a type of the workflow, determining a group of historical metrics associated with the type of the workflow, using a model generated based on machine learning applied to the group of historical metrics, determining respective predicted ranks for different settings that are able to be applied to the workflow, based on the respective predicted ranks, determining a setting, of the different settings, to use to apply to the workflow and that has a predicted rank of the respective predicted ranks that satisfies a defined criterion, and applying the setting to at least one of the group of files to decrease a latency associated with processing the workflow.

Abstract: Systems and methods are provided for enhanced pose generation based on generative modeling. An example method includes accessing an autoencoder trained based on poses of real-world persons, each pose being defined based on location information associated with joints, with the autoencoder being trained to map an input pose to a feature encoding associated with a latent feature space. Information identifying, at least, a first pose and a second pose associated with a character configured for inclusion in an in-game world is obtained via user input, with each of the poses being defined based on location information associated with the joints and with the joints being included on a skeleton associated with the character. Feature encodings associated with the first pose and the second pose are generated based on the autoencoder. Output poses are generated based on transition information associated with the first pose and the second pose.

Abstract: A method is provided for operating a multi-channel messaging system. The method may provide automated conversation across multiple communication channels associated with a user by: selecting a first communication channel from the multiple communication channels, wherein the first communication channel is hosted by a first communication server; receiving, via an existing user interface of the first communication channel, a first user input via as part of a conversation with a chatbot, wherein the chatbot comprises a communication data structure comprising a plurality of communication paths; selecting a communication path based on the first user input; and generating a first feedback in response to the first user input according to the selected communication path.

Abstract: The present disclosure relates to a feature contribution system that accurately and efficiently provides the influence of features utilized in machine-learning models with respect to observed model results. In particular, the feature contribution system can utilize an observed model result, initial contribution values, and historical feature values to determine a contribution value correction factor. Further, the feature contribution system can apply the correction factor to the initial contribution values to determine correction-factor adjusted contribution values of each feature of the model with respect to the observed model result.

Abstract: More natural communication and interaction are enabled. The autonomous system (100) includes an action decision unit (140) that decides, based on an attention level map (40) in which an attention level indicating the degree of attention for each position in a predetermined space is set, the action which a drive mechanism is caused to perform.

Abstract: A dialog system and intelligent personal assistant capable of semantic-understanding mapping between user intents and machine services includes an interface to receive a voice or text request from a user. Semantics of the request including at least one of entities, intent, or context are extracted from the user's request. A sequence of action features is selected based on the extracted semantics. A sequence of application programming interfaces (APIs) corresponding to the sequence of action features is then executed to generate a result. An action sequence optimizer may optimize the sequence of action features based on user configuration. The examples provide a technical solution to model a richer semantic-understanding mapping between user intents and available APIs, that will be greatly improving user experience in the spoken dialogue system, as core of personal assistant.

Abstract: Systems, computer program products, and methods are described herein for mitigating data loss in an edge computing environment using machine learning and distributed ledger techniques. The present invention is configured to receive an indication that one or more edge computing nodes is processing one or more portions of a transaction; retrieve dark data associated with each of the one or more edge computing nodes; initiate a machine learning algorithm on the dark data retrieved from each of the one or more edge computing nodes; capture, using the machine learning algorithm, representation information for each of the one or more edge computing nodes from their respective dark data; generate a ledger record for the representation information for each of the one or more edge computing nodes; and register the ledger record for the representation information for each of the one or more edge computing nodes on a first distributed ledger.

Abstract: Systems and methods for similarity search are described. Embodiments identify a document and a query corresponding to a matching phrase in the document, encode the query and a candidate phrase, score the candidate phrase using at least one learning-based score and at least one surface form score, wherein the at least one learning based score is based on the encoding, and the at least one surface form score is based on a surface form of the query and a surface form of the candidate phrase, and select the matching phrase based on the scoring.

Abstract: An online system generates a set of genomic representations, each including multiple genes, in which each gene represents users assigned to a control or test group for performing a test. A metric is identified based on a treatment associated with the test group and a score for each representation is computed based on a difference between two values, in which each value is based on the metric associated with users assigned to the test or control group. A propagation process is executed by identifying representations having at least a threshold score, propagating genes included in the representations to an additional set of representations through recombination and/or mutation, and computing the score for each additional representation. The propagation process is repeated for each additional set of representations until stopping criteria are met and a representation is selected based on scores associated with one or more representations.

Abstract: Aspects of the disclosure relate to computing platforms that implement data driven testing through automation using continuous machine learning combined with continuous integration for development testing. A computing platform may extract production data from user interaction with an application, perform a masking operation on the production data to create modified production data, convert the modified production data into a format suitable for machine learning, and generate one or more independent scenarios using machine learning. The computing platform may automatically trigger execution of test scripts associated with the one or more independent scenarios using the modified production data and receive results of the executed test scripts indicating whether a version of the application performs as expected.

Abstract: Systems and methods are disclosed for managing a generative artificial intelligence (AI) model to improve performance. Managing the generative AI model includes using a second generative AI model to generate outputs from similar inputs and comparing the outputs of the generative AI models to determine their similarity. The second generative AI model is trained using fresher training data but may not be manually supervised and adjusted to the level of the generative AI model being managed. As such, an output of the second generative AI model is compared to an output of the managed generative AI model by a classification model to determine a relevance of the output from the managed generative AI model. An output of the classification model is used to perform various suitable policies to optimize the performance of the managed generative AI model, such as providing alternate outputs, preventing providing the output, or retraining the model.

Abstract: Techniques and embodiments are described herein for detecting and mitigating fraudulent activity within user experience (UX) test applications. In some embodiments, a system applies a set of rules and/or machine learning (ML) models to each respondent of an online survey or UX test. Different ML models may be trained to learn domain-specific patterns indicative of fraudulent activity. The system may then select the ML models based on attributes of the UX test and/or respondent. The selected rules and/or ML models may generate a probabilistic score representing a likelihood that the respondent is currently engaging in or will engage in fraudulent activity with respect to a UX test. If the score exceeds a threshold, then the system may take action to mitigate the fraudulent activity, such as triggering the removal of the user from an accepted respondent pool, halting further engagement between the respondent and the UX test, and generating alerts.

Abstract: Automating conversion of drawings to indoor maps and plans. One example is a computer-implemented method of machine learning, the method comprising: receiving, by a device, a first set of updates to a first indoor map, the first indoor map previously created by a production machine-learning system having a production map accuracy; training, by a device, a supporting machine-learning system using the first set of updates to the first indoor map; and then applying, by a device, test data to the supporting machine-learning system, the applying results in first-evaluation indoor map with a first-evaluation map accuracy; and when the first-evaluation map accuracy is within a predetermined window above the production map accuracy refraining, by a device, from updating the production machine-learning system based on the first set of updates.

Abstract: Systems and methods for retrieval of augmented parameters for an artificial intelligence (AI)-based character are provided. An example method includes receiving, from a user via a user interface, at least one keyword describing the AI-based character; retrieving, from at least one data source and based on the at least one keyword, the augmented parameters describing the AI-based character; and generating, based on the augmented parameters, an AI-based character model corresponding to the AI-based character. The at least one data source includes a database configured to store records associated with the AI-based character, an online search service, and a set of clusters associated with a type of a feature of the AI-based character and at least one hidden prompt corresponding to the type of the feature. The type of the feature includes one of the following: a voice, a dialog style, an emotional state, an age, and temperament.

Abstract: The present disclosure provides systems and methods that perform machine-learned natural language processing. A computing system can include a machine-learned natural language processing model that includes an encoder model trained to receive a natural language text body and output a knowledge graph and a programmer model trained to receive a natural language question and output a program. The computing system can include a computer-readable medium storing instructions that, when executed, cause the processor to perform operations. The operations can include obtaining the natural language text body, inputting the natural language text body into the encoder model, receiving, as an output of the encoder model, the knowledge graph, obtaining the natural language question, inputting the natural language question into the programmer model, receiving the program as an output of the programmer model, and executing the program on the knowledge graph to produce an answer to the natural language question.

Abstract: A method is provided for operating a multi-channel messaging system. The method may provide automated conversation across multiple communication channels associated with a user by: selecting a first communication channel from the multiple communication channels, wherein the first communication channel is hosted by a first communication server; receiving, via an existing user interface of the first communication channel, a first user input via as part of a conversation with a chatbot, wherein the chatbot comprises a communication data structure comprising a plurality of communication paths; selecting a communication path based on the first user input; and generating a first feedback in response to the first user input according to the selected communication path.

Abstract: Disclosed is a dynamic network risk predicting method based on a graph neural network, which comprises the following steps: collecting network data firstly and constructing a network data sequence graph; extracting the time sequence of the network data sequence graph and obtaining a time sequence feature graph; extracting network attribute features, network structure features and network change features of the dynamic network; obtaining the representation vector of the network sequence graph by learning the representation vector; constructing the anomaly detection model and carrying out the dynamic network risk prediction.

Abstract: Methods and systems are provided for sequence-to-sequence conversion from unstructured search queries to structured database queries, so that lay persons may retrieve information from relational databases without specialized knowledge of database query languages. An encoder module and a decoder module of a learning model are trained to convert an unstructured search query to an intermediate feature vector by computing co-attention and self-attention based on a context string and a database schema, encoding the database schema in the context string by application of self-attention between the context string containing tokens of the database schema with learned structural attention heads which relate the token to logic of the database. Training is performed using labeled training datasets which include structured database queries which are normalized by parsing into a semantic representation thereof, followed by linearization.

Abstract: Provided is a cluster node recommendation system. A method of controlling the cluster node recommendation system includes: inputting user selection information from a user, the user selection information including at least one of a cloud vendor, an Information Technology (IT) resource size, and a free resource size; checking resource requirements of a designated application; outputting a node configuration by inputting the input user selection information and the checked resource requirements of the application to an artificial intelligence module; verifying validity by arranging a container in which the application is executed, in the output node configuration; and providing a final node configuration in which validity verification is made, to the user.

Abstract: The present invention may have a computer that monitors conversation between a user and a first virtual assistant from a plurality of virtual assistants, for a confidence score value being below a confidence threshold value, where the confidence score value is associated with each answer of the first virtual assistant. When the computer determines the confidence score value is below the confidence threshold value, the computer generates a dialog flow for answering by a second virtual assistant, where the dialog flow is related to an answer associated with the confidence score value being below the confidence threshold value. Then, based on determining the second virtual assistant having a second confidence score value that is higher than the confidence score value, the computer transfers a knowledge base data of a topic associated with the dialog flow from the first virtual assistant to the second virtual assistant.

Abstract: Method and system are provided for uncertainty quantification analysis with efficient two dimensional (2D) random fields. The method may be used in a modeling system. The method may include providing a covariance matrix of a set of parameters with a covariance function describing an occurring real world phenomenon; and generating a random field based on the covariance matrix including a controlled rank reduction to produce a reduced-rank random field dataset provided as a data structure with data compression. Generating the random field applies rank reduction of a block circulant with circulant blocks (BCCB) representation of a covariance matrix obtained from the covariance matrix using existing symmetry of eigenvalues to eliminate redundant computations. The method may include outputting a reduced-rank random field dataset for use in a user-provided model of the modeling system for uncertainty quantification analysis and simulation of the real world phenomenon.

Abstract: A fault identification may be triggered by a component of a power generation system (PGS), such as a hardware component, a controller of a hardware component, a device of the PGS, a computer connected to the PGS, a computer configured to monitor the PGS, and/or the like. The fault identification may be the result of a failure of a component of the PGS, a future failure of a component of the PGS, a routine maintenance of the PGS, and/or the like. The fault is converted to a notification on a user interface using a mapping of faults, root-causes, notification rules, and/or the like. The conversion may use one or more lookup tables and/or formulas for determining the impact of the fault on the PGS, and/or the like.

Abstract: A system and method for managing medication orders, including changes to those orders, is described. A server receives medication orders from a pharmacy and communicates those orders to an infusion pump. The infusion pump communicates status messages to the server and the server compares the status messages to the medication orders to determine if the there were any changes to the order, or if a medication was administered absent a medication order. The server assigns a new order identification number to the status message if needed. The system may also reconcile original medication orders with information associated with a new order identification number and stores the information in the eMAR of a patient.

Abstract: Systems and methods for labeling data are disclosed. An example method may be performed by one or more processors of a labeling system and include retrieving labeled data, identifying characteristics predictive of labels that would be entered for unlabeled data items having the respective characteristics based on the labeled data, training an analysis model to predict labels that would be entered for unlabeled data items, generating, for unlabeled data items, using the trained analysis model, a prediction of a label that will be entered for the respective unlabeled data item if the respective unlabeled data item is presented for labeling, selecting, based on the generated predictions, a subset of unlabeled data items to be presented for labeling, receiving labels for the subset of unlabeled data items, determining that a completion criteria associated with the trained analysis model is met, and generating labels for remaining unlabeled data items.

Abstract: A system for responding to external requests received at an internal bot network is provided. The system may include an internal bot network including a plurality of internal bots. The plurality of internal bots may interact with a plurality of external bots. The system may include a translation layer/barrier. Each bot included in the plurality of internal bots and the plurality of external bots may be resident on one or more hardware processors. Each external bot may communicate using its own unique set of external specifications. Each internal bot may communicate using a universal set of internal specifications and is prevented from communicating using a set of external specifications. The translation layer/barrier may intercept requests and responses between the plurality of internal bots and the plurality of external bots. The translation layer/barrier reformats intercepted requests and responses to correspond to the set of specifications specific to the receiving bot.

Abstract: The described technology is generally directed towards a machine learning development hub, and corresponding methods and computer readable media. The machine learning development hub can comprise a machine learning development platform complete with various tools for various stages of machine learning development. The machine learning development hub can furthermore comprise translation functions to translate received inputs into inputs to other machine learning development platforms. The machine learning development hub can collect credentials for the other machine learning development platforms and can connect to the other machine learning development platforms via their respective interfaces, in order to supply inputs and instructions thereto. The machine learning development hub can encrypt its communications to other machine learning development platforms to secure its interactions.

Abstract: An example computing platform is configured to (i) cause a client device to display an interface for deploying a new data science environment, where the interface presents (a) a list of data science applications and (b) a set of user-defined configuration parameters, (ii) receive, from the client device, data indicating (a) a user selection of a given data science application from the list and (b) a user selection of one or more user-defined configuration parameters from the set, (iii) based on the user selection of the given data science application, determine a deployment template for use in deploying the new data science environment, the deployment template specifying (a) an executable environment package and (b) a set of predefined configuration parameters, and (iv) use the given executable environment package, the set of predefined configuration parameters, and the one or more user-defined configuration parameters to deploy the new data science environment.

Abstract: An approach is provided in which the approach generates a parallel chart based on multiple records that includes a set of variable values corresponding to a set of variables. To generate the parallel chart, the approach arranges the set of variables on the parallel chart in a variable order based on at least one variable arrangement rule. The approach arranges an initial variable value order for each one of the set of variables, and computes a lucidity score based on the variable order and the initial variable value order of each of the set of variables. The lucidity score is a measurement of the clarity of the parallel chart. The approach adjusts the variable value order of at least one of the set of variables to increase the lucidity score and optimizes the clarity of the parallel chart based on the adjusted variable value order.

Abstract: An apparatus, computer-readable medium, and computer-implemented method for efficiently classifying a data object, including representing the data object as a data object vector in a vector space, each dimension of the data object vector corresponding to a different feature of the data object, determining a distance between the data object vector and centroids of data domain clusters in the vector space, each data domain cluster comprising data domain vectors representing data domains, sorting the data domain clusters according to their respective distances to the data object vector, and iteratively applying data domain classifiers corresponding to data domains represented in a closest data domain cluster in the sorted data domain clusters to the data object.

Abstract: Techniques for tuning parameters for machine learning models are described. Different values for a parameter are tested to determine the value that results in an optimized model. A parameter value may be selected for testing using a search algorithm based on how the model performs with respect to other values for the parameter. Different values may be tested until a stopping criterion (such as time for testing, number of trials, amount of enhancement in performance, etc.) is met. In some embodiments, the techniques may be used to determine parameter values for natural language processing models.

Abstract: An electronic device and method are provided. The method includes identifying a speech section of a user and a speech section of a neighbor in a received audio signal, identifying a user utterance in the speech section of the user and a neighbor answer to the user utterance in the speech section of the neighbor, obtaining preference information associated with the user utterance, giving a first reliability to the neighbor answer and a second reliability to an agent answer of an artificial intelligence agent generated in response to the user utterance, based on the preference information, not responding to the user utterance when the second reliability is lower than the first reliability, and outputting the agent answer when the second reliability is equal to or higher than the first reliability.

Abstract: Methods, computer program products, and systems are presented. The methods include, for instance: creating a log data specific lexicon based on log data samples, each word in the log data specific lexicon corresponding to a weighted sentiment score with a binary polarity. A log message is obtained, and a sentiment value of the log message is assigned based on respective weighted sentiment scores of words appearing in the log message. The log message is classified for a class indicating an issue the log message addresses. An alert type for the log message is determined based on the sentiment value, the class, and a priority of the log message, and a system alert including a set of key performance indicators according to the alert type is produced to a user.

Abstract: Broadly speaking, the present techniques relate to a method and apparatus for performing action recognition, and in particular to a computer-implemented method for performing action recognition on resource-constrained or lightweight devices such as smartphones. The ML model may be adjusted to achieve required accuracy and efficiency levels, while also taking into account the computational capability of the apparatus that is being used to implement the ML model. One way is to adjust the number of channels assigned to the first set of channels, i.e. the full temporal resolution channels. Another way is to adjust the point in the ML model where the temporal pooling layer or layers are applied.