# US 8,751,416 B2 · Method and apparatus for deriving probabilistic models from deterministic ones

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 8,751,416 B2
- Document ID: US8751416B2
- Status: Granted
- Kind: B2
- Application number: 12/439,617
- Filed: 2007-08-28
- Published: 2010-01-21
- Granted: 2014-06-10
- Earliest claimed: 2006-08-28
- Assignees: Koninklijke Philips N.V.
- Applicants: Nicolas W. Chbat; Kees Van Zon
- Inventors: Nicolas W. Chbat; Kees Van Zon
- Web page: https://patentgrep.com/patent/US8751416B2

## CPC

- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/00: Machine learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A Dynamic Bayesian Network provides models that provides emulation of patient data.

## Claims

### Claim 1

A system for deriving a probabilistic model from a deterministic model, comprising:
a validated deterministic model (VDM), wherein the VDM is a dynamic Bayesian Network (DBN) which is derived from a system of ordinary differential equations, the DBN being provided as a file representing variables and parameters, probabilistic relationships of the variables and strengths of the probabilistic relationships as defined by the parameters; and
a machine learning algorithm (MLA) operative to receive data from the VDM and to generate the probabilistic model, wherein the probabilistic model receives the data, selects an appropriate class of parameters for the DBN, and emulates a patient according to a patient class.

### Claim 2 (depends on 1)

A system as claimed in claim 1, wherein the MLA further receives other parameters.

### Claim 3 (depends on 1)

A system as claimed in claim 1, wherein the probabilistic model receives input variables and provides output variables during deployment.

### Claim 4

A system for emulating variables in a person, comprising:
a validated deterministic model (VDM) operative to provide patient data, wherein the VDM is a dynamic Bayesian Network (DBN) which is derived from a system of ordinary differential equations, the DBN being provided as a file representing variables and parameters, probabilistic relationships of the variables and strengths of the probabilistic relationships as defined by the parameters; and
a machine learning algorithm (MLA) operative to receive patient data from the VDM and to generate the probabilistic model, wherein input variables are provided to the probabilistic model to provide the emulating variables and wherein the probabilistic model receives the patient data, selects an appropriate class of parameters for the DBN, and emulates a patient according to a patient class.

### Claim 5 (depends on 4)

A system as claimed in claim 4, wherein the MLA further receives other parameters.

### Claim 6 (depends on 4)

A system as claimed in claim 4, wherein the input variables provided to the probabilistic model are measured patient data.

### Claim 7 (depends on 4)

A system as claimed in claim 4, wherein the MLA receives parameters based on patient information.

### Claim 8 (depends on 7)

A system as claimed in claim 7, wherein the patient information includes patient class-specific data.