Secure messaging in a machine learning blockchain network
Multi-layer ensembles of neural subnetworks are disclosed. Implementations can classify inputs indicating various anomalous sensed conditions into probabilistic anomalies using an anomaly subnetwork. Determined probabilistic anomalies are classified into remedial application triggers invoked to recommend or take actions to remediate, and/or report the anomaly. Implementations can select a report type to submit, or a report recipient, based upon the situation state, e.g., FDA: Field Alert Report (FAR), Biological Product Deviation Report (BPDR), Medwatch, voluntary reporting by healthcare professionals, consumers, and patients (Forms 3500, 3500A, 3500B, Reportable Food Registry, Vaccine Adverse Event Reporting System (VAERS), Investigative Drug/Gene Research Study Adverse Event Reports, Potential Tobacco Product Violations Reporting (Form 3779), USDA: APHIS Center for Veterinary Biologics Reports, Animal and Plant Health Inspection Service: Adverse Event Reporting, FSIS Electronic Consumer Complaints, DEA Tips, Animal Drug Safety Reporting, Consumer Product Safety Commission Reports, State/local reports: Health Department, Board of Pharmacy.
1. A method of finding reportable adverse events to be reported from an input data set including parameters describing a first event involving a first physical object, including:
receiving by a trusted server having access permissions to a server network, a request by an unverified source to report an adverse event, the report identifying the first physical object;
establishing by the trusted server, a trusted interactive conversation with the unverified source, the trusted interactive conversation identified by a unique token, wherein physical parameters of the first physical object are prompted for, and values for the physical parameters are returned;
using a machine intelligence process, determining by the trusted server from the values for the physical parameters received, values for one or more associated event parameters;
creating by the trusted server, using the values for one or more associated event parameters, a new report identifying the first physical object and populated with event parameters determined by the machine intelligence process to be associated with the adverse event; and
notifying other trusted servers with access to the server network, of the new report as populated.
2. The method of claim 1 , wherein the machine intelligence process further includes determining values for one or more associated event parameters using a trained neural network classifier trained using a training set of example physical parameter values and example associated event parameters.
3. The method of claim 1 , wherein the machine intelligence process further includes:
applying a semi-supervised statistical machine learner to documents stored with the server network to develop patterns of data recognized from within the documents;
using the semi-supervised statistical machine learner to encode data representing healthy states (Xh) and unhealthy states (Xu) to latent vectors representing event level data mapped into a latent vector space; and
using the semi-supervised statistical machine learner to identify data points, corresponding to events of a certain Mahalanobis distance from data clusters labelled as desirable.
4. The method of claim 3 , further including identifying patterns in the data points identified, the patterns indicating clusters of data, and applying a label to each cluster; and continuously training a classifier including a neural network with at least some of the clusters and the labels.
5. The method of claim 1 , wherein values for one or more associated event parameters are determined by aggregating information obtained from multiple reports of events in the server network selected by the machine intelligence process using the physical parameters.
6. The method of claim 1 , wherein values for one or more associated event parameters are determined from a set including at least a plurality of counterfeit, diverted, stolen, intentional adulteration, unfit for distribution, and fraudulent activity.
7. The method of claim 1 , wherein values for physical parameters are detected from a set including at least a plurality of a presence of an adulterating substance, such as metallic or glass filings, medication color, a marking on a pill, foul or repugnant odors, an empty vial, a discolored contents, a cracked vial, a damaged packaging, a precipitated medicine inside, and a mislabeled vial.
8. The method of claim 1 , wherein first event involving a first physical object is selected from a set of a missing pill bottle, a mis-delivery of a therapeutic, a mis-delivery of a medical service delivered, a mis-coding of a therapeutic or a service properly delivered, an adulterated substance, a delivery not made, and a patient sick or dead in connection with a product, or an adverse reaction event in connection with a product problem.
9. The method of claim 1 , further including the trusted server generating and providing a unique random token to the unverified source, accompanied by a link to a trusted application for making interactive requests for information of the unverified source, the link linking to a copy of the trusted application stored at an application store for download, the trusted application verifying the trusted server to the unverified source.
10. The method of claim 1 , further including obtaining from other trusted servers with access to the server network, indication of additional instances of like events based upon the values for one or more associated event parameters in the new report as populated.
11. A non-transitory computer readable medium storing instructions for finding reportable events to be reported from an input data set including parameters describing a first event involving a first physical object, which instructions when executed by one or more processors perform a method comprising:
receiving by a trusted server having access permissions to a server network, a request by an unverified source to report an adverse event, the report identifying the first physical object;
establishing by the trusted server, a trusted interactive conversation with the unverified source, the trusted interactive conversation identified by a unique token, wherein physical parameters of the first physical object are prompted for, and values for the physical parameters are returned;
using a machine intelligence process, determining by the trusted server from the values for the physical parameters received, values for one or more associated event parameters;
creating by the trusted server, using the values for one or more associated event parameters, a new report identifying the first physical object and populated with event parameters determined by the machine intelligence process to be associated with the adverse event; and
notifying other trusted servers with access to the network, of the new report as populated.
12. The non-transitory computer readable medium of claim 11 , wherein the machine intelligence process further includes determining associated event parameters using a trained neural network classifier trained using a training set of example physical parameter values and example associated event parameters.
13. The non-transitory computer readable medium of claim 11 , wherein the machine intelligence process further includes:
applying a semi-supervised statistical machine learner to documents stored with the server network to develop patterns of data recognized from within the documents;
using the semi-supervised statistical machine learner to encode data representing healthy states (Xh) and unhealthy states (Xu) to latent vectors representing event level data mapped into a latent vector space; and
using the semi-supervised statistical machine learner to identify data points corresponding to events of a certain Mahalanobis distance from data clusters labelled as desirable.
14. The non-transitory computer readable medium of claim 13 , further including identifying patterns in the data points indicating clusters of data, and applying a label to each cluster; and continuously training a classifier including a neural network with at least some of the clusters and the labels.
15. The non-transitory computer readable medium of claim 11 , wherein values for one or more associated event parameters are determined by aggregating information obtained from multiple reports of events in the server network selected by the machine intelligence process using the physical parameters.
16. The non-transitory computer readable medium of claim 11 , wherein values for one or more associated event parameters are determined from a set including at least a plurality of counterfeit, diverted, stolen, intentional adulteration, unfit for distribution, and fraudulent activity.
17. The non-transitory computer readable medium of claim 11 , wherein values for physical parameters are detected from a set including at least a plurality of a presence of an adulterating substance, such as metallic or glass filings, medication color, a marking on a pill, foul or repugnant odors, an empty vial, a discolored contents, a cracked vial, a damaged packaging, a precipitated medicine inside, and a mislabeled vial.
18. The non-transitory computer readable medium of claim 11 , wherein first event involving a first physical object is selected from a set of a missing pill bottle, a mis-delivery of a therapeutic, a mis-delivery of a medical service delivered, a mis-coding of a therapeutic or a service properly delivered, an adulterated substance, a delivery not made, and a patient sick or dead in connection with a product, or an adverse reaction event in connection with a product problem.
19. The non-transitory computer readable medium of claim 11 , further including the trusted server generating and providing a unique random token to the unverified source, accompanied by a link to a trusted application for making interactive requests for information of the unverified source; wherein the link links to a copy of the trusted application stored at an application store for download, the trusted application verifying the trusted server to the unverified source.
20. The non-transitory computer readable medium of claim 11 , further including obtaining from other trusted servers with access to the server network, indication of additional instances of like events based upon the values for one or more associated event parameters in the new report as populated.
21. A system for finding reportable events to be reported from an input data set including parameters describing a first event involving a first physical object, the system comprising:
an interface to a trusted source system for publishing to a server of the trusted source system a mobile application being provided to one or more non-credentialed devices;
a trusted server having access to a plurality of trusted server nodes storing at least pointers to a plurality of documents, the trusted server including a processor and software that implement:
receiving by a trusted server having access permissions to a server network, a request by an unverified source to report an adverse event, the report identifying the first physical object;
establishing by the trusted server using the interface to a trusted source system for publishing to a server of the trusted source system a mobile application being provided to the non-credentialed devices, a trusted interactive conversation with the unverified source, the trusted interactive conversation identified by a unique token, wherein physical parameters of the first physical object are prompted for, and values for the physical parameters are returned;
using a machine intelligence process, determining by the trusted server from the values for the physical parameters received, values for one or more associated event parameters;
creating by the trusted server, using the values for one or more associated event parameters, a new report identifying the first physical object and populated with event parameters determined by the machine intelligence process to be associated with the adverse event; and
notifying other trusted servers with access to the server network, of the new report as populated.