Method for finding solutions

• US 6,768,973 B1
• Filed: 10/19/2000
• Issued: 07/27/2004
• Est. Priority Date: 04/12/2000
• Status: Active Grant

• Current Assignee: Unilever Home & Personal Care Usa Division Of Conopco, Inc.
• Original Assignee: Unilever Home & Personal Care Usa Division Of Conopco, Inc.
• Inventors: Patel, Shail
• Primary Examiner(s): PALADINI, ALBERT WILLIAM
• Application Number:
  US09/691,803
• Time in Patent Office: 1,377 Days
• Field of Search: 703/2, 705/7, 705/8, 706/21, 706/19, 700/28, 700/FOR.102
• US Class Current: 703/2
• CPC Class Codes: G06Q 10/04 Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" market predictions or forecasting for commercial activities G06Q30/0202
• CPC Class Codes: G06Q 10/06375 Prediction of business process outcome or impact based on a proposed change

Abstract

A method is described for using a computer in finding solutions to a problem dependent on a number of criteria. A model algorithm is provided for each of the criteria, each model algorithm providing a prediction for a corresponding criteria when a candidate solution is inputted into the model algorithm. At least one model algorithm provides a prediction with a prediction error bar for the corresponding criteria. Criteria are selected to optimise a set of candidate solutions. An algorithm is provided for optimisation of the set of candidate solutions in accordance with the selected criteria. A first set of candidate solutions is provided and the optimisation algorithm generates one or more new candidate solutions. All candidate solutions are inputted into the number of model algorithms to obtain predictions and at least one prediction error bar. Information of the set of candidate solutions obtained by the generation and/or previous optimisations and/or experiments is used to select candidate solutions from the set to obtain an optimised set of candidate solutions.

Claims

1. A computer implemented method for finding solutions to a problem dependent on a number of criteria, comprising the steps of

i providing a model algorithm for each of the criteria, each model algorithm providing a prediction for a corresponding criteria when a candidate solution is inputted into the model algorithm and wherein at least one model algorithm provides a prediction with a prediction error bar for the corresponding criteria;

ii selecting criteria to optimise a set of candidate solutions; and

iii providing an algorithm for optimisation of the set of candidate solutions in accordance with the selected criteria;

wherein a first set of candidate solutions is provided,

wherein the optimisation algorithm generates one or more new candidate solutions, wherein all candidate solutions are inputted into the number of model algorithms to obtain predictions and at least one prediction error bar, and wherein information of the set of candidate solutions obtained by said generation and/or previous optimisations and/or experiments is used to select candidate solutions from the set to obtain an optimised set of candidate solutions.

2. A computer implemented method according to claim 1, including a number of iteration steps in a loop as follows:

(i) a first set of one or more candidate formulations are used as a starting point;

(ii) candidate solutions are inputted into the number of model algorithms to obtain predictions whereby at least one prediction includes a prediction error bar; and

(iii) the optimisation algorithm generates new candidate solutions; and

(iv) the new candidate formulations are used as input into the number of model algorithms in iteration step (ii); and

wherein predictions and prediction error bars of the set of candidate solutions are used to select candidate solutions from the set to obtain an optimised set of candidate solutions.

3. A computer implemented method according to claim 1, wherein the optimisation algorithm generates new candidate solutions using information of the generated candidate solutions to improve the set of candidate solutions, the information comprising solutions, predictions and prediction error bars, the gradients and estimated gradients of predictions and/or prediction error bars, the constraints, and optionally any other, information available.

4. A computer implemented method according to claim 1, wherein constraints are defined regarding the candidate solutions and/or the predictions and/or the prediction error bar(s), wherein these constraints are used in the optimisation algorithm in generating the candidate solutions.

5. A computer implemented method according to claim 1, wherein candidate solutions are selected comprising predictions with minimum prediction error bars.

6. A computer implemented method according to claim 1, wherein a weighted sum(s) of prediction(s) and prediction error bar(s) of candidate solutions is(are) determined for selecting the candidate solutions.

7. A computer implemented method according to claim 1, wherein Pareto optimal sets of predictions and prediction error bars of candidate solutions are determined to select candidate solutions.

8. A computer implemented method according to claim 1, wherein the candidate solutions are displayed against selected sets of two of said number of criteria.

10. A computer implemented method according to claim 1, wherein constraints on the criteria can be introduced in an interactive manner to filter the optimised set of candidate solutions.

11. A computer implemented method according to claim 1, wherein information of the optimised set of candidate solutions is used to determine a region of an experimental space for carrying out further experiments.

12. A computer implemented method according to claim 11, wherein the results of the further experiments are used to improve one or more of the model algorithms.

13. A computer implemented method according to claim 1, wherein a Bayesian neural network algorithm is used as a model algorithm providing a prediction and prediction error bar.

14. A computer implemented method according to claim 1, wherein the first set of candidate solutions is generated in a random manner, is seeded from previous actual solutions and/or previous candidate solutions or optimised sets of candidate solutions.

15. A computer implemented method according to claim 14, wherein candidate solutions for the first set of solutions are obtained by the method of claim 1, wherein the weighted optimisation algorithm is used.

16. A computer implemented method according to claim 1, wherein the method is used in formulation optimisation.

17. A computer implemented method according to claim 1, wherein the method is used in optimising manufacturing processes.

18. A computer program device readable by a computer, comprising a computer program executable by the computer for effecting the computer to carry out the method of claim 1.

19. A computer program in a format downloadable by a computer, comprising a computer program executable by the computer to install the program in the computer for execution to effect the computer to carry out the method of claim 1.

9. A computer implemented method according to claim((s)) 8, wherein constraints on specific criteria can be introduced during displaying the candidate solutions against these specific criteria.