Abstract: The present embodiments relate to a language identification system for predicting a language and text content of text lines in an image-based document. The language identification system uses a trainable neural network model that integrates multiple neural network models in a single unified end-to-end trainable architecture. A CNN and an RNN of the model can process text lines and derive visual and contextual features of the text lines. The derived features can be used to predict a language and text content for the text line. The CNN and the RNN can be jointly trained by determining losses based on the predicted language and content and corresponding language labels and text labels for each text line.

Abstract: A document repository may be searched for documents that are similar to a source document. Multiple queries may be generated based on a type of the source document, and the results may be combined in a unified response. User behavior may then be monitored, and implicit and explicit feedback may be gathered to evaluate the performance of the search. The gathered feedback may indicate how relevant each of the result documents are in comparison to the original source document. This feedback may then be used to adjust search parameters for the source document type, such that the performance of subsequent searches may be improved. A model may also be trained to classify implicit feedback using explicit feedback received from users.

Abstract: Routines are used to connect bot queries entered via a bot interface and database queries executed on a database. Each routine is associated with routine configurations, including (a) query attributes associated with bot queries that trigger the routine, (2) a database query executed or to be executed for the routine, and/or (3) display settings for displaying information returned from the database query in a bot response. Each routine is configured to generate an output structured data object (SDO) indicating information returned from the database query and the display settings applicable to the information. The output SDOs from the various routines are transmitted to the bot application via a single API endpoint. The routine configurations are entered by a design user through one or more design user interfaces rendered by an integrative configuration application module.

Abstract: Techniques are disclosed for creating an attachment between two compute instances. An infrastructure and a generalized method is described for attaching two or more cloud resources (e.g., two compute instances) in spite of the compute resources being provisioned by two different services from different cloud tenancies. An automated process is described that is executed for wiring the compute instances. The automated process can be generally applied to attach any two compute instances providing two different services and provisioned from two different service tenancies.

Abstract: Techniques for modifying rules for completing requests using a conversational interface are disclosed. A conversational interface receives a user request to perform a particular function. A system extracts semantic content from the request and obtains contextual metadata associated with the request. Using the semantic content and contextual metadata, the system generates a data set of attribute values. The system determines whether the data set matches a data set of required data attributes associated with the request. If the provided data set is different than the required data set, the system determines whether to modify a set of rules associated with the request. The set of rules specifies which attribute values are required to complete the request. When the system modifies the rules associated with the request, the system also determines whether to apply the rule modification to subsequent requests by the same user and other users.

Abstract: A document may be received as part of a request to identify similar documents in a collection of documents. However, the received document and the documents in the collection may have different schemas or formats. To provide semantic context to the search and allow similarity scores to be generated between different document types, a configuration may be accessed that defines how to generate queries from one schema into another schema. The configuration may map queries between different fields in both schemas. Results of the multiple queries can be combined to generate a weighted combination for each document that can be used as a similarity score between different document types.

Abstract: The present embodiments relate to identifying a native language of text included in an image-based document. A cloud infrastructure node (e.g., one or more interconnected computing devices implementing a cloud infrastructure) can utilize one or more deep learning models to identify a language of an image-based document (e.g., a scanned document) that is formed of pixels. The cloud infrastructure node can detect text lines that are bounded by bounding boxes in the document, determine a primary script classification of the text in the document, and derive a primary language for the document. Various document management tasks can be performed responsive to determining the language, such as perform optical character recognition (OCR) or derive insights into the text.

Abstract: Techniques for recommending runbook operations during a runbook generation process are disclosed. A system recommends operations for including in the runbook based on attributes of a detected event. The system presents event attributes associated with the detected event to a user in a runbook generation interface. When the user selects an event attribute, the system presents a set of candidate runbook operations associated with the event attribute. Based on user selections of runbook operations to include in a runbook, the system generates and stores a template of the runbook in which the attribute values associated with the detected event are omitted. Upon execution of the runbook at a future time, the system, or a user, may populate the event attributes in the runbook with values that correspond to the future event of the same event type.

Abstract: Techniques for collaboratively executing runbook sessions are disclosed. A system provides for collaborative completion of runbook operations by delegating runbook operations or by transferring ownership of a runbook from one user to another. A system identifies a first set of runbook operations capable of being completed by a first user. The system identifies a second set of runbook operations that are not capable of being completed by the first user. The system identifies a second user capable of completing the second set of runbook operations. Upon completion of the first set of runbook operations by the first user, the system determines whether to delegate the second set of runbook operations to the second user or to transfer ownership of the runbook session to the second user. The system determines that the runbook session is completed based on the combined results collaboratively generated by the initial user and the alternate user.

Abstract: Techniques are disclosed for using a trained machine learning model to generate a career progression pathways that are evaluated in view of employment conditions and compromises (trade-offs) that are acceptable to an employee. The system trains the machine learning model using employee profiles. The employee profiles include employment histories, skills, credentials, and professional activities. Once trained, the system applies the machine learning model to an employee's profile to generate ML-based career progression paths for reach a target employment goal. Each ML-based career progression path defines one or more interim objectives for reaching the target employment goal. The system compares the interim objectives, as defined by the ML-based career progression paths, with new employment conditions that are acceptable to an employee. The system recommends a subset of the ML-based career progression path(s) with interim objectives that are compatible with the acceptable employment conditions.

Abstract: Techniques are disclosed for restricting operations between two attached two compute instances. An infrastructure and a generalized method is described for attaching two or more cloud resources (e.g., two compute instances) in spite of the compute resources being provisioned by two different services from different cloud tenancies, and then modifying the allowed operations that can be performed due to the attachment.

Abstract: Techniques for recommending runbook operations during a runbook generation process are disclosed. The system recommends operations for including in the runbook based on user activity. Generating operations of a runbook based on user activity generates a runbook that captures a user's behavior. The recommended operations may be different than any operations performed by the user. The recommended operations may be based in part on data sets accessed by the user frequently, periodically, or continuously.

Abstract: The present embodiments relate to identifying and mitigating memory bit flips in a cloud infrastructure service. The cloud infrastructure service can provide a monitoring system to monitor low level memory space to detect bit flips by the DRAM instances in the cloud infrastructure service. The bit flips detected in various DRAM computing instances can be processed to verify that the bit flips are sustained (e.g., and possibly relating to a Rowhammer attack) rather than transitory bit flips occurring in DRAM computing devices. Responsive to validating a set of bit flips at one or more computing instances, workloads associated with the affected computing instances can be migrated to other computing instances in the cloud infrastructure service.

Abstract: Techniques for providing intelligent data profiling services are disclosed. A distributed computing system can include a data profiling service and a data recommender service. During an interactive data session, a user can select a data column of a data entity and request data profiling information for the data column. The data profiling service can compute the requested profile. The data recommender service can then determine a priority rank for the data columns of the data entity. Based on the priority rank, the data profiling service can compute a second data profile for a second data column of the data entity. The second data profile can be stored in a cache for rapid retrieval when the user requests the second data profile.

Abstract: Techniques for generating supplemental information based on runbook operation results are disclosed. A system generates and displays supplemental information for a runbook execution interface based on one of a system component associated with an executable operation of a runbook, and a set of runbook operation results corresponding to the executable operation. The system receives a user input to execute an operation defined by a runbook presented to remediate an event. The system generates supplemental information for the runbook execution interface based on the results of the operation executed by the user. The system identifies characteristics associated with the runbook operation results and identifies sources for additional information. Source may include performance data from the same component over a different period of time, performance data of a similar component, and performance data of topologically-connected components.

Abstract: Embodiments relate to configuring artificial-intelligence (AI) decision nodes throughout a communication decision tree. The decision nodes can support successive iteration of AI models to dynamically define iteration data that corresponds to a trajectory through the tree.

Abstract: Techniques for identifying a root cause of an operational result of a deterministic machine learning model are disclosed. A system applies a deterministic machine learning model to a set of data to generate an operational result, such as a prediction of a “fault” or “no-fault” in the system. The set of data includes signals from multiple different data sources, such as sensors. The system applies an abductive model, generated based on the deterministic machine learning model, to the operational result. The abductive model identifies a particular set of data sources that is associated with the root cause of the operational result. The system generates a human-understandable explanation for the operational result based on the identified root cause.

Abstract: Techniques for improving a semantic parser of a dialog system, by breaking the semantic parser into a coarse semantic parser and a fine semantic parser, are described. A method described herein includes accessing an utterance received in a dialog system. The utterance is a text-based natural language expression. The method further includes applying a coarse semantic parser to the utterance to determine an intermediate logical form for the utterance. The intermediate logical form indicates one or more intents in the utterance. The method further includes applying a fine semantic parser to the intermediate logical form to determine a logical form for the utterance. The logical form is a syntactic expression of the utterance according to an established grammar, and the logical form includes one or more parameters of the one or more intents. The logical form can be used to conduct a dialog with a user of the dialog system.

Abstract: Techniques for selecting a runbook to as a recommendation to diagnose or remediate an event are disclosed. A system identifies an event, such as an anomaly in monitored metrics values. The system identifies runbooks associated with the event. The system analyzes the runbooks to determine which runbooks to recommend for remediating the event. The analysis includes previewing results of operations associated with steps of the runbooks to identify interesting results. The system performs a runbook operation prior to recommending the runbook for diagnosing or remediating the event. Based on the results of the operations, the system determines a relevance of a runbook to the diagnosis or remediation of the detected event. The system recommends a runbook for diagnosing or remediating the event based on the relevance of the set of results to the diagnosis or remediation of the event.

Abstract: Techniques for selecting a runbook to recommend for remediating a detected event are disclosed. When a system detects an event, the system obtains metadata associated with the event. The metadata provides information about the event and a system topology of the system in which the event occurred. The system generates a recommendation for a runbook to remediate the event based on one or both of characteristics of the event and characteristics of the topology in which the event occurred. The system compares a system topology to system topologies associated with previously-executed runbooks. The system recommends one of the previously-executed runbooks to remediate a detected event based on determining that the topology associated with the previously-executed runbook is similar to the topology of the system in which the event occurred.

Abstract: A drift analysis system (DAS) is described that is capable of automatically detecting potential model schema drift issues when a machine learning model (MIL model), which has been trained using a particular training dataset, is used to make a prediction for a particular input provided to the model. The DAS performs one or more drift checks by comparing characteristics of the input to characteristics of the training dataset that was used to train the model that is being used to make a prediction for the input. Results obtained by the DAS from performing the drift checks may then be output along with the prediction made for the particular input. The one or more drift check results may be compiled into a drift report, which may be served concurrently with prediction results generated by the trained machine-learning model for the input.

Abstract: Embodiments implement non-intrusive load monitoring using machine learning. A trained convolutional neural network (CNN) can be stored, where the CNN includes a plurality of layers, and the CNN is trained to predict disaggregated target device energy usage data from within source location energy usage data based on training data including labeled energy usage data from a plurality of source locations. Input data can be received including energy usage data at a source location over a period of time. Disaggregated target device energy usage can be predicted, using the trained CNN, based on the input data.

Abstract: Herein are techniques that concurrently populate entries in a compressed sparse row (CSR) encoding, of a type of edge of a heterogenous graph. In an embodiment, a computer obtains a mapping of a relational schema to a graph data model. The relational schema defines vertex tables that correspond to vertex types in the graph data model, and edge tables that correspond to edge types in the graph data model. Each edge type is associated with a source vertex type and a target vertex type. For each vertex type, a sequence of persistent identifiers of vertices is obtained. Based on the mapping and for a CSR representation of each edge type, a source array is populated that, for a same vertex ordering as the sequence of persistent identifiers for the source vertex type, is based on counts of edges of the edge type that originate from vertices of the source vertex type.

Abstract: The present disclosure relates to systems and methods for providing cloud-based services securely to on-premises networks or other infrastructure. More particularly, the present disclosure relates to systems and methods for enriching first-party data (e.g., data collected directly by an on-premises server) stored within on-premises networks by enabling the on-premises networks to retrieve and process third-party data stored on cloud-based networks. As a technical benefit, cloud-based services can be performed on the first-party data within the on-premises networks.

Abstract: A data object has a lock and a condition indicator associated with it. Based at least partly on detecting a first setting of the condition indicator, a reader stores an indication that the reader has obtained read access to the data object in an element of a readers structure and reads the data object without acquiring the lock. A writer detects the first setting and replaces it with a second setting, indicating that the lock is to be acquired by readers before reading the data object. Prior to performing a write on the data object, the writer verifies that one or more elements of the readers structure have been cleared.

Abstract: In accordance with an embodiment, described herein are systems and methods for enabling multiple parents with weights in a multidimensional database. An exemplary system can comprise a computer that includes one or more microprocessors, and a multidimensional database server executing on the computer, wherein the multidimensional database server supports at least one hierarchical structure of data dimensions. A data dimension can comprises a plurality of members, a first set of members being a first level, and a second set of members being at a second level. A member of the second set of members at the second level can comprises two or more member-member relationships with two or more members of the first set of members at the first level. Each of the two or more member-member relationships are associated with a weight.

Abstract: Described herein are systems and methods for suppressing inbound payload to an integration flow of an orchestration based application integration. The systems and methods described herein can, based upon a scan of an integration, identify and exclude from memory certain portions of one or more payloads that are received at the integration flow.

Abstract: A method for analyzing software with pointer analysis may include obtaining a software program, and determining a first independent program slice of the software program describing a first code segment of the software program. The method may further include determining, using a first pointer analysis objective, a first result from performing a first pointer analysis on the first independent program slice, and determining, using the first result, a first dependent program slice of the software program. The method may further include determining, using a second pointer analysis objective, a second result from performing a second pointer analysis on the first dependent program slice. The method may further include generating a report, using these results, indicating whether the software program satisfies a predetermined criterion.

Abstract: Embodiments of the invention include techniques for storing and managing repositories of electronic presentation documents. Presentation storage repositories may be implemented in various storage systems including individual computer servers, distributed storage networks, cloud-based storage systems, and the like. Presentation search interfaces may support various searching and customization capabilities. Search interface and search engine components may be implemented to receive search requests from client devices, retrieve electronic presentation data from one or more storage repositories, and generate customized search results based on the retrieved electronic presentation data. Search and presentation capabilities may include storing user-specific selections of presentation cover slides, and then retrieving and/or generating user-specific thumbnails for search results interfaces.

Abstract: Techniques are disclosed for generating an execution plan for performing functional tests in a cloud-computing environment. Infrastructure resources and capabilities (e.g., system requirements) may be defined within an infrastructure object (e.g., a resource of a declarative infrastructure provisioner) that stores a code segment that implements the resource or capability. Metadata may be maintained that indicates what particular capabilities are applicable to each infrastructure resource. Using the metadata, the system can generate an execution plan by combining code segments for each resource with code segments defining each capability in accordance with the metadata. The execution plan may include programmatic instructions that, when executed, generate a set of test results. The system can execute instructions that cause the set of test results to be presented at a user device.

Abstract: Adaptive data collections may include various type of data arrays, sets, bags, maps, and other data structures. A simple interface for each adaptive collection may provide access via a unified API to adaptive implementations of the collection. A single adaptive data collection may include multiple, different adaptive implementations. A system configured to implement adaptive data collections may include the ability to adaptively select between various implementations, either manually or automatically, and to map a given workload to differing hardware configurations. Additionally, hardware resource needs of different configurations may be predicted from a small number of workload measurements. Adaptive data collections may provide language interoperability, such as by leveraging runtime compilation to build adaptive data collections and to compile and optimize implementation code and user code together.

Abstract: The embodiments disclosed herein relate to using machine learning to allocate a number of concurrent processes for minimizing the completion time for executing a task having multiple subtasks. Historical data comprising a variety of subtask types with actual completion times is mined to create a set of statistical models for predicting completion time for a type of subtask. To minimize the total time to complete execution of a new task, a certain number of threads is allocated to execute subtasks of the new task. The certain number of threads is determined based on the predicted completion time for the subtasks using the respective statistical model. Threads are assigned to subtasks based on the predicted completion time for the subtasks, and the subtasks assigned to each thread are scheduled for execution.

Abstract: Systems and methods for namespace masking in an integration flow. In order to mitigate issues arising due to application updates impacting integration flows, during the metadata generation phase, all the elements which are available for operation selection can be cloned and replaced with a new namespace. The original elements are deleted from the metadata. The systems and methods can maintain the original namespace and the element name at a predefined place in the metadata (e.g., as an attribute of the cloned Element). The cloned element in addition to all the original fields of the element, can have one additional field which will store the value of the original namespace.

Abstract: Techniques discussed herein relate to providing composable edge devices. In some embodiments, a user request specifying a set of services to be executed at a cloud-computing edge device may be received by a computing device operated by a cloud computing provider. A manifest may be generated in accordance with the user request. The manifest may specify a configuration for the cloud-computing edge device. Another request can be received specifying the same or a different set of services to be executed at another edge device. Another manifest which specifies the configuration for that edge device may be generated and subsequently used to provision the request set of services on that device. In this manner, manifests can be used to compose the platform to be utilized at any given edge device.

Abstract: The present disclosure describes techniques for using instances of a connector to communicate with multiple target systems without a manual pre-configuration step for each of the multiple target systems. The connector may be an extensible program-code-template for creating objects to perform one or more operations. For example, each instance of the connector may obtain information regarding a corresponding target system, determine an interface to the corresponding target system based on the information, and translate requests for the corresponding target system according to the interface.

Abstract: Techniques are described herein for enabling autonomous agents to generate conclusive answers. An example of a conclusive answer is text that addresses concerns of a user who is interacting with an autonomous agent. For example, an autonomous agent interacts with a user device, answering user utterances, for example questions or concerns. Based on the interactions, the autonomous agent determines that a conclusive answer is appropriate. The autonomous agent formulates the conclusive answer, which addresses multiple user utterances. The conclusive answer provided to the user device.

Abstract: Embodiments relate to configuring artificial-intelligence (AI) decision nodes throughout a communication decision tree.

Abstract: Techniques for providing request throttling using proportional, integral, and exponential smoothing algorithms are disclosed. A distributed computing system can include a throttler engine that receives a plurality of requests targeting a software component within the distributed computing system. The throttler engine can aggregate the requests into a queue based on a time window. The throttler engine can determine a received request rate and a request rate limit for the software component and then compute a throttled request rate. The throttled request rate can include correction terms derived from proportional and integral computations and a correction term obtained from an exponential smoothing algorithm. The throttler engine can then provide throttled requests from the queue to the software component.

Abstract: Systems, devices, and methods discussed herein are directed to utilizing patterns and logical entities to identify and maintain relationships between data assets. In some embodiments, a query comprising a logical entity qualifier, one or more pattern identifiers that indicate a pattern, and a data set identifier may be received. The pattern is executed against a data set corresponding to the data set identifier and one or more logical entities are generated based on this execution. A logical entity may be a label that represents a set of one or more data assets in a data set. Assets that share a label can share attributes that are described by the label. The label corresponding to each logical entity may be presented, where each label represents a different set of data assets which share a common trait. In some embodiments, the user may define a pattern by which commonality may be assessed.

Abstract: Techniques are disclosed for generating an execution plan for performing functional tests in a cloud-computing environment. Infrastructure resources and capabilities (e.g., system requirements) may be defined within an infrastructure object (e.g., a resource of a declarative infrastructure provisioner) that stores a code segment that implements the resource or capability. Metadata may be maintained that indicates what particular capabilities are applicable to each infrastructure resource. Using the metadata, the system can generate an execution plan by combining code segments for each resource with code segments defining each capability in accordance with the metadata. The execution plan may include programmatic instructions that, when executed, generate a set of test results. The system can execute instructions that cause the set of test results to be presented at a user device.

Abstract: Techniques for representing a native function using an executable reference are disclosed. The system receives an instruction to create an executable reference for a native function, including a method type comprising a method signature corresponding to the executable reference, and a function description including (a) a memory layout corresponding to data returned by the function and (b) memory layouts corresponding to parameters required by the function. The system selects an application binary interface (ABI). The system generates code that, for each parameter, of the one or more parameters required by the function, converts the parameter from a value formatted for use by a Java Virtual machine to a value formatted for use in the native function, based on the selected ABI. Responsive to invocation of the executable reference, the generated code and the native function may be executed.

Abstract: Disclosed techniques relate to tracing a subsidiary thread of a web page application. In an example, a method includes providing a web page application to a web browser. The method further includes creating, from the web page application, a global tracer configured to log tracing data for the web page application and a span. The method further includes instantiating, from the web page application, a wrapper for a subsidiary thread. The wrapper is configured to execute shim code before executing the subsidiary thread. The method further includes passing, from the wrapper to the shim code, configuration data of the global tracer. The method further includes creating a subsidiary tracer from the shim code and the configuration data. The subsidiary tracer configured to log tracing data for the subsidiary thread and a thread span.

Abstract: Systems, devices, and methods of the present disclosure use discourse analysis and other techniques to form questions and answers from text. The questions and answers can be used for different applications, including providing a virtual dialogue or generating training data for machine-learning models. For example, a dialogue application generates a discourse tree that represents text and identifies a question from a satellite elementary discourse unit of the discourse tree. The dialogue application annotates the text by inserting the generated question and labeling the satellite elementary discourse unit as an answer.

Abstract: Techniques for time-series analysis for forecasting computational workloads are disclosed. A resource management system monitors a computing system and obtains metrics data from the computing system. The metrics data is stored as a set of historical data points of a data set. A first portion of the set of historical data points comprise an outlier that does not correspond to a seasonality pattern associated with a second portion of the set of historical data points. The resource management system tests a first time-series model that incorporates a first exogenous variable corresponding to a first exogeneous factor to determine that the first time-series model fits both the first portion of the set of historical data points and the second portion of the set of historical data points within an error threshold. Then, the resource management system selects the first time-series model to predict future data points of the data set.

Abstract: Distributed ledgered data is stored within a distributed persistent storage system comprising multiple persistent storage systems as distributed ledgered participants. In various embodiments, the distributed ledgered data is maintained using the native capabilities of a persistent storage system. The distributed ledgered data is replicated as persistent data objects in a “ledgered repository of objects” that are replicated at each of the persistent storage systems. Changes at one persistent storage system are recorded within a block in a distributed blockchain that is distributed across each of the other distributed ledgered participants. The other distributed ledgered participants read the changes from the blockchain and apply the changes to the respective replicas at each of the other distributed ledgered participants. Hence, this approach is referred to as blockchain apply. Blockchain apply may be used to replicate the repository objects of various forms of PSSs.

Abstract: A method for executing a second-order taint analysis on library code may include generating, by executing a first-order taint analysis on the library code starting at a sink, a first execution path from a load instruction to the sink. The load instruction may perform: reading a first value using a first global identifier. The method may further include determining a store instruction by matching the load instruction and the store instruction. The store instruction may perform: writing a second value using a second global identifier. The method may further include, generating a second execution path from the store instruction to the load instruction, generating, by executing the first-order taint analysis on the library code starting at the store instruction, a third execution path from an entry point to the store instruction, and forming a potential second-order taint flow by joining the first, second, and third execution paths.

Abstract: The disclosed embodiments provide a system that estimates a remaining useful life (RUL) for a fan. During operation, the system receives telemetry data associated with the fan during operation of the critical asset, wherein the telemetry data includes a fan-speed signal. Next, the system uses the telemetry data to construct a historical fan-speed profile, which indicates a cumulative time that the fan has operated in specific ranges of fan speeds. The system then computes an RUL for the fan based on the historical fan-speed profile and empirical time-to-failure (TTF) data, which indicates a TTF for the same type of fan as a function of fan speed. Finally, when the RUL falls below a threshold, the system generates a notification indicating that the fan needs to be replaced.

Abstract: Techniques are described herein for executing queries on distinct portions of a database object that has been separate into chunks and distributed across the volatile memories of a plurality of nodes in a clustered database system. The techniques involve redistributing the in-memory database object portions on changes to the clustered database system. Each node may maintain a mapping indicating which nodes in the clustered database system store which chunks, and timestamps indicating when each mapping entry was created or updated. A query coordinator may use the timestamps to select a database server instance with local in memory access to data required by a portion of a query to process that portion of the query.

Abstract: A method for selecting multiple network function (NF) types using a single discovery request includes receiving, by a NF repository function (NRF) from a consumer NF, a single discovery request message that requests discovery of a plurality of producer NF types in a fifth generation (5G) core network and includes an indication of multiple NF discovery requests, obtaining, by the NRF, NF profile objects corresponding to each of the plurality of producer NF types included in the single discovery request message, and providing, to the consumer NF via a single discovery response message generated by the NRF, NF profile objects that correspond to one or more of the plurality of producer NF types.

Abstract: A model-based approach to determining an optimal configuration for a data center may use an environmental chamber to characterize the performance of various data center configurations at different combinations of temperature and altitude. Telemetry data may be recorded from different configurations as they execute a stress workload at each temperature/altitude combination, and the telemetry data may be used to train a corresponding library of models. When a new data center is being configured, the temperature/altitude of the new data center may be used to select a pre-trained model from a similar temperature/altitude. Performance of the current configuration can be compared to the performance of the model, and if the model performs better, a new configuration based on the model may be used as an optimal configuration for the data center.