Abstract: In one embodiment, a method for auditing the results of a machine learning model includes: retrieving a set of state estimates for original time series data values from a database under audit; reversing the state estimation computation for each of the state estimates to produce reconstituted time series data values for each of the state estimates; retrieving the original time series data values from the database under audit; comparing the original time series data values pairwise with the reconstituted time series data values to determine whether the original time series and reconstituted time series match; and generating a signal that the database under audit (i) has not been modified where the original time series and reconstituted time series match, and (ii) has been modified where the original time series and reconstituted time series do not match.

Abstract: Systems, methods, and other embodiments associated with associated with dependency checking for machine learning (ML) models are described. In one embodiment, a method includes applying a repeating probe signal to an input signal input into a machine learning model. An estimate signal output from the machine learning model is monitored, and the repeating probe signal is checked for in the estimate signal. Based on the results of the checking for the repeating probe signal, an evaluation of dependency in the machine learning model is presented.

Abstract: We describe a system that performs prognostic-surveillance operations based on an inferential model that dynamically adapts to evolving operational characteristics of a monitored asset. During a surveillance mode, the system receives a set of time-series signals gathered from sensors in the monitored asset. Next, the system uses an inferential model to generate estimated values for the set of time-series signals, and then performs a pairwise differencing operation between actual values and the estimated values for the set of time-series signals to produce residuals. Next, the system performs a sequential probability ratio test (SPRT) on the residuals to produce SPRT alarms. When a tripping frequency of the SPRT alarms exceeds a threshold value, which is indicative of an incipient anomaly in the monitored asset, the system triggers an alert. While the prognostic-surveillance system is operating in the surveillance mode, the system incrementally updates the inferential model based on the time-series signals.

Abstract: The disclosed embodiments relate to a system that automatically adapts a prognostic-surveillance system to account for aging phenomena in a monitored system. During operation, the prognostic-surveillance system is operated in a surveillance mode, wherein a trained inferential model is used to analyze time-series signals from the monitored system to detect incipient anomalies. During the surveillance mode, the system periodically calculates a reward/cost metric associated with updating the trained inferential model. When the reward/cost metric exceeds a threshold, the system swaps the trained inferential model with an updated inferential model, which is trained to account for aging phenomena in the monitored system.

Abstract: The disclosed embodiments relate to a system that optimizes a prognostic-surveillance system to achieve a user-selectable functional objective. During operation, the system allows a user to select a functional objective to be optimized from a set of functional objectives for the prognostic-surveillance system. Next, the system optimizes the selected functional objective by performing Monte Carlo simulations, which vary operational parameters for the prognostic-surveillance system while the prognostic-surveillance system operates on synthesized signals, to determine optimal values for the operational parameters that optimize the selected functional objective.

Abstract: The disclosed embodiments relate to a system that automatically adapts a prognostic-surveillance system to account for aging phenomena in a monitored system. During operation, the prognostic-surveillance system is operated in a surveillance mode, wherein a trained inferential model is used to analyze time-series signals from the monitored system to detect incipient anomalies. During the surveillance mode, the system periodically calculates a reward/cost metric associated with updating the trained inferential model. When the reward/cost metric exceeds a threshold, the system swaps the trained inferential model with an updated inferential model, which is trained to account for aging phenomena in the monitored system.

Abstract: In one embodiment, a method for auditing the results of a machine learning model includes: retrieving a set of state estimates for original time series data values from a database under audit; reversing the state estimation computation for each of the state estimates to produce reconstituted time series data values for each of the state estimates; retrieving the original time series data values from the database under audit; comparing the original time series data values pairwise with the reconstituted time series data values to determine whether the original time series and reconstituted time series match; and generating a signal that the database under audit (i) has not been modified where the original time series and reconstituted time series match, and (ii) has been modified where the original time series and reconstituted time series do not match.

Abstract: We describe a system that performs prognostic-surveillance operations based on an inferential model that dynamically adapts to evolving operational characteristics of a monitored asset. During a surveillance mode, the system receives a set of time-series signals gathered from sensors in the monitored asset. Next, the system uses an inferential model to generate estimated values for the set of time-series signals, and then performs a pairwise differencing operation between actual values and the estimated values for the set of time-series signals to produce residuals. Next, the system performs a sequential probability ratio test (SPRT) on the residuals to produce SPRT alarms. When a tripping frequency of the SPRT alarms exceeds a threshold value, which is indicative of an incipient anomaly in the monitored asset, the system triggers an alert. While the prognostic-surveillance system is operating in the surveillance mode, the system incrementally updates the inferential model based on the time-series signals.

Abstract: In a knowledge classification system, both the information sources and queries are processed to generate knowledge representation graph structures. The graph structures for both the query and the information sources are then converted to views and displayed to a searcher. By manipulating the graph structure views for each information source, the searcher can examine the source for relevance. A search can be performed by comparing the graph structure of the query to the graph structure of each information source by a graph matching computer algorithm. Information sources are classified by constructing hierarchies of knowledge representations. The simplest construction is obtained by using the knowledge representation of a query as the top of the hierarchy. The structures in the hierarchy are substructures of the query. The hierarchy of structures may also be constructed by using the knowledge representation of the query as the bottom of the hierarchy.

Abstract: In a knowledge classification system, both the information sources and queries are processed to generate knowledge representation graph structures. The graph structures for both the query and the information sources are then converted to views and displayed to a searcher. By manipulating the graph structure views for each information source, the searcher can examine the source for relevance. A search can be performed by comparing the graph structure of the query to the graph structure of each information source by a graph matching computer algorithm. Information sources are classified by constructing hierarchies of knowledge representations. The simplest construction is obtained by using the knowledge representation of a query as the top of the hierarchy. The structures in the hierarchy are substructures of the query. The hierarchy of structures may also be constructed by using the knowledge representation of the query as the bottom of the hierarchy.

Abstract: In a knowledge classification system, both the information sources and queries are processed to generate knowledge representation graph structures. The graph structures for both the query and the information sources are then converted to views and displayed to a searcher. By manipulating the graph structure views for each information source, the searcher can examine the source for relevance. A search can be performed by comparing the graph structure of the query to the graph structure of each information source by a graph matching computer algorithm. Information sources are classified by constructing hierarchies of knowledge representations. The simplest construction is obtained by using the knowledge representation of a query as the top of the hierarchy. The structures in the hierarchy are substructures of the query. The hierarchy of structures may also be constructed by using the knowledge representation of the query as the bottom of the hierarchy.

Abstract: A distributed computer database system connected to a network, e.g., the Internet or on an intranet, indexes interests of agents that have registered with the system, examines information objects, for example, that reside on the network, and, responsive to a match with the registered agents' interests, specifies to the agents the relevant information objects.

Abstract: A distributed computer database system includes one or more front end computers, one or more home nodes, one or more index nodes and one or more object nodes interconnected by a network into a search engine for retrieval of hypertext documents. A query from a user is transmitted to one of the front end computers, which forwards the query to one of the home nodes, of the search engine. The home node parses the query into one or more elementary queries and schedules the elementary queries for processing. Each elementary query can be one of a number of types, including an index query, a link query or an object query. To process an index query or link query, the home node extracts features from the index query or link query, fragments the extracted features into feature fragments, and hashes these features. Each hashed feature fragment is transmitted to one index node on the network.

Abstract: An information retrieval apparatus for processing a query for retrieval of information from a database has a mechanism for locating a number of features and feature fragments in an index database; an evaluating mechanism for identifying a number of sub-queries of a number of levels contained in the query and recursively evaluating the sub-queries using each of the located features and feature fragments; and a mechanism for collecting and storing a number of results of the recursive evaluation of the query and sub-queries pursuant to computing an overall result of the query. Such a system can eliminate the need of conventional retrieval systems for providing a new, separate, centralized replica within the data warehouse of the data stored in the diverse external databases. The invention can thus avoid the problems of replication of such data in conventional systems, in which the data may become stale or is subject to errors arising during replication for warehousing.

Abstract: A distributed computer database system including one or more front end computers and one or more computer nodes interconnected by a network into a search engine for retrieval of objects including images, sound and video streams, as well as plain and structured text. A query is an object in the same format as the objects to be retrieved. A query from a user is transmitted to one of the front end computers which forwards the query to one of the computer nodes, termed the home node, of the search engine. The home node extracts features from the query and hashes these features. Each hashed feature is transmitted to one node on the network. Each node on the network which receives a hashed feature uses the hashed feature of the query to perform a search on its respective partition of the database. The results of the searches of the local databases are gathered by the home node.

Abstract: A distributed computer database system includes one or more front end computers and one or more computer nodes interconnected by a network into a search engine for retrieval of objects processed by a variety of interrelated ontologies. Each object conforms to a specific ontology. A query is an object which conforms to a specific ontology, which is to be used for retrieval of objects conforming to one or more target ontologies. A query from a user is transmitted to one of the front end computers which forwards the query to one of the computer nodes, termed the home node, of the search engine. The home node extracts features from the query, according to its ontology. These features are then hashed. Each hashed feature and the list of target ontologies is transmitted to one node on the network. Each node on the network which receives a hashed feature uses the hashed feature of the query to perform a search on its respective partition of the database.

Abstract: A distributed computer database system connected to a network, e.g., the Internet or on an intranet, indexes interests of agents that have registered with the system, examines information objects, for example, that reside on the network, and, responsive to a match with the registered agents' interests, specifies to the agents the relevant information objects.

Abstract: A distributed computer database system including a front end computer and a plurality of computer nodes interconnected by a network into a search engine. A query from a user is transmitted to the front end computer which forwards the query to one of the computer nodes, termed the home node, of the search engine. The home node fragments the query and hashes the fragments of query to create an index by which the hashed query fragments are transmitted to one or more nodes on the network. Each node on the network which receives a hashed fragment uses the fragment of the query to perform a search on its respective database. The results of the searches of the local databases are then gathered by the home node.