Abstract: The present disclosure relates to systems, methods, and products for optimization of a chromatography purification process using a physics-informed neural network. The method includes inputting a plurality of process parameters into the physics-informed neural network to obtain a predicted output; calculating a loss function based on a set of governing equations, as set of constraints, and the predicted output; determining whether the physics-informed neural network is convergent based on the calculated loss function; in response to the physics-informed neural network being convergent, exporting the physics-informed neural network; and in response to the physics-informed neural network not being convergent: updating a plurality of weights in the physics-informed neural network, and inputting the plurality of process parameters to the physics-informed neural network for a next convergence iteration to calculate the loss function and determine whether the physics-informed neural network is convergent.

Abstract: A system maintains a knowledge layout to support the building of event response recommendations. Meta-graph patterns may be used to determine semantic relatedness between events and actions in response. Event-action node pairs are then constructed.

Abstract: A device may receive a request to identify items that satisfy parameters of the request. The device may identify a plurality of items that satisfy the parameters. The device may generate a plurality of explanation sets. An explanation set of the plurality of explanation sets may relate to an item of the plurality of items. The explanation set may include at least one of: a positive explanation indicating that the item is positively associated with a first characteristic that relates to a first preference of a user, or a negative explanation indicating that the item is negatively associated with a second characteristic that relates to a second preference of the user. The device may select an item from the plurality of items based on the plurality of explanation sets. The device may provide information that includes an explanation set of the item selected.

Abstract: The present disclosure relates to systems, methods, and products for adaptive design and optimization using a physics-informed neural network (PINN). The system includes a non-transitory memory and a processor. The processor executes instructions to cause the system to: input collocation points and design parameters into the PINN to obtain an output; calculate a loss function based on a set of governing equations and the output; determine whether the PINN is convergent based on the calculated loss function; in response to the PINN being convergent, export the PINN; and in response to the PINN not being convergent: determine whether to resample the collocation points; determine an optimum number of collocation points; determine a set of optimal network parameters for adjusting the PINN; and input the collocation points and the set of optimal network parameters to the PINN for a next iteration.

Abstract: A systems implements a gradient descent calculation, regression calculation, or other machine learning calculation on a dataset (e.g., a global dataset) using a coordination node including coordination circuitry that coordinates multiple worker nodes to create a distributed calculation architecture. In some cases, the worker nodes each hold a portion of the dataset and operate on their respective portion. In some cases, the gradient descent calculation, regression calculation, or other machine learning calculation is used to implement a targeted maximum likelihood scheme for causal inference estimation. The targeted maximum likelihood scheme may be used to conduct causal analysis of the observational data.

Abstract: The present disclosure relates to a system, a method, and a product for optimizing hyper-parameters for generation and execution of a machine-learning model under constraints. The system includes a memory storing instructions and a processor in communication with the memory. When executed by the processor, the instructions cause the processor to obtain input data and an initial hyper-parameter set; for an iteration, to build a machine learning model based on the hyper-parameter set, evaluate the machine learning model based on the target data to obtain a performance metrics set, and determine whether the performance metrics set satisfies the stopping criteria set. If yes, the instructions cause the processor to perform an exploitation process to obtain an optimal hyper-parameter set, and exit the iteration; if no, perform an exploration process to obtain a next hyper-parameter set, and perform a next iteration with using the next hyper-parameter set as the hyper-parameter set.

Abstract: A causal inference stack implements a targeted maximum likelihood scheme to conduct causal analysis of the observational data. At a data-handling layer, the causal inference stack obtains one or more memory locations for a dataset and establishes analysis nodes to setup localized data handling for the memory locations. At a data classification layer, the causal inference stack characterizes the missingness of the dataset. At a pipeline layer, the causal inference stack obtains a data element dependency query from a user and sets up an end-to-end solution path to determine the presence of a causal relationship between data elements identified in the data element dependency query.

Abstract: In some implementations, a control system may obtain historical data associated with usage of a distillation column during a historical time period. The control system may configure a prediction model to monitor the distillation column for a hazardous condition. The prediction model may be trained based on training data that is associated with occurrences of the hazardous condition. The control system may monitor, using the prediction model, the distillation column to determine a probability that the distillation column experiences the hazardous condition within a threshold time period. The prediction model may be configured to determine the probability based on measurements from a set of sensors of the distillation column. The control system may perform, based on the probability satisfying a probability threshold, an action associated with the distillation column to reduce a likelihood that the distillation column experiences the hazardous condition within the threshold time period.

Abstract: A resource prediction system for executing machine learning models and method are provided. The system includes non-transitory memory storing instructions and a processor configured to execute the instructions to obtain input data including a targeted objective and the constraints, select a deployable machine learning model having an evaluation score that meets a predetermined criterion from among candidate machine learning models, virtually execute the deployable machine learning model on each of candidate hardware platforms according to the constraints, generate an assessment report of the virtual performance metrics set of the deployable machine learning model executed on each of the candidate hardware platforms, and select the suggested hardware platform meeting the predetermined criterion from among the candidate hardware platforms.

Abstract: The present disclosure describes a system, a method, and a product for computational resource prediction of user tasks and subsequent workload provisioning. The computational resource predictions for a user task is achieved using a twin machine learning and AI system based on probabilistic programing. The workload scheduling and assignment of the user task in a computing cluster with components having diverse hardware architectures are further managed by an automatic and intelligent assignment/provisioning engine based on various machine learning and AI models and reinforcement learning. The automatic workload scheduling and assignment engine is further configured to handle unpredicted uncertainty and adapt to constantly evolving system queues of the tasks submitted by the users to generate queuing/re-queuing, running/termination, and resource allocation/reallocation actions for user tasks.

Abstract: In some implementations, a control system may obtain historical data associated with usage of a distillation column during a historical time period. The control system may configure a prediction model to monitor the distillation column for a hazardous condition. The prediction model may be trained based on training data that is associated with occurrences of the hazardous condition. The control system may monitor, using the prediction model, the distillation column to determine a probability that the distillation column experiences the hazardous condition within a threshold time period. The prediction model may be configured to determine the probability based on measurements from a set of sensors of the distillation column. The control system may perform, based on the probability satisfying a probability threshold, an action associated with the distillation column to reduce a likelihood that the distillation column experiences the hazardous condition within the threshold time period.

Abstract: Complex computer system architectures are described for detecting a concept drift of a machine learning model in a production environment, for adaptive optimization of the concept drift detection, for extracting embedded features associated with the concept drift using a shadow learner, and for adaptive adjustment of the machine learning model in production to mitigate the effect of predictive performance drop due to the concept drift.

Abstract: This disclosure is directed to a generalizable machine learning model production environment and system with a defense mechanism that facilitates safe execution of machine learning models in production by effectively detecting potential known and new adversarial attacks. The disclosed exemplary systems and architectures gather data from the online execution of the machine learning models and communicate with an on-demand pipelines for further inspection and/or correction of vulnerabilities in the production machine learning model to the detected attacks. These systems and architectures provide an automatable process for continuous monitoring of model performance and correction of the production machine learning model to guard against current and future adversarial attacks.

Abstract: A system maintains a knowledge layout to support the building of event response recommendations. Meta-graph patterns may be used to determine semantic relatedness between events and actions in response. Event-action node pairs are then constructed.

Abstract: The present disclosure relates to a system, a method, and a product for optimizing hyper-parameters for generation and execution of a machine-learning model under constraints. The system includes a memory storing instructions and a processor in communication with the memory. When executed by the processor, the instructions cause the processor to obtain input data and an initial hyper-parameter set; for an iteration, to build a machine learning model based on the hyper-parameter set, evaluate the machine learning model based on the target data to obtain a performance metrics set, and determine whether the performance metrics set satisfies the stopping criteria set. If yes, the instructions cause the processor to perform an exploitation process to obtain an optimal hyper-parameter set, and exit the iteration; if no, perform an exploration process to obtain a next hyper-parameter set, and perform a next iteration with using the next hyper-parameter set as the hyper-parameter set.

Abstract: A systems implements a gradient descent calculation, regression calculation, or other machine learning calculation on a dataset (e.g., a global dataset) using a coordination node including coordination circuitry that coordinates multiple worker nodes to create a distributed calculation architecture. In some cases, the worker nodes each hold a portion of the dataset and operate on their respective portion. In some cases, the gradient descent calculation, regression calculation, or other machine learning calculation is used to implement a targeted maximum likelihood scheme for causal inference estimation. The targeted maximum likelihood scheme may be used to conduct causal analysis of the observational data.

Abstract: Complex computer system architectures are described for providing a machine learning model management tool that monitors, detects, and makes revisions to machine learning models to prevent declines and maintain robustness and fairness in machine learning model performance in production over time. The machine learning model management tool achieves its goals via intelligent management, organization, and orchestration of detection, inspection, and correction engines.

Abstract: A device may receive a request to identify items that satisfy parameters of the request. The device may identify a plurality of items that satisfy the parameters. The device may generate a plurality of explanation sets. An explanation set of the plurality of explanation sets may relate to an item of the plurality of items. The explanation set may include at least one of: a positive explanation indicating that the item is positively associated with a first characteristic that relates to a first preference of a user, or a negative explanation indicating that the item is negatively associated with a second characteristic that relates to a second preference of the user. The device may select an item from the plurality of items based on the plurality of explanation sets. The device may provide information that includes an explanation set of the item selected.