Adaptive network for automated first break picking of seismic refraction events and method of operating the same
First Claim
1. A method of operating an adaptive network to set values of a plurality of weighting factors therein, comprising:
- storing data representing a sample input and a desired output in memory;
retrieving a portion of said stored data representing said sample input;
operating the adaptive network, using the data retrieved in said retrieving step representing the sample input as a data input to the adaptive network, to produce an actual output;
calculating a sample error, by determining the difference between the actual output of said adaptive network and the desired output;
comparing the sample error to a tolerance limit;
responsive to the result of the comparing step indicating that the sample error exceeds the tolerance limit, calculating a learning factor according to the value of the sample error, so that said learning factor increases as the sample error decreases;
adjusting at least one of the plurality of weighting factors in said adaptive network by backpropagation according to the generalized delta rule, using the sample error and the learning factor;
repeating the retrieving and presenting steps, and the step of calculating the sample error, until said comparing step indicates that the sample error does not exceed the tolerance limit.
0 Assignments
0 Petitions
Accused Products
Abstract
A method of operating an adaptive, or neural, network is disclosed for performing first break analysis for seismic shot records. The adaptive network is first trained according to the generalized delta rule. The disclosed training method includes selection of the seismic trace with the highest error, where the backpropagation is performed according to the error of this worst trace. The learning and momentum factors in the generalized delta rule are adjusted according to the value of the worst error, so that the learning and momentum factors increase as the error decreases. The training method further includes detection of slow convergence regions, and methods for escaping such regions including restoration of previously trimmed dormant links, renormalization of the weighting factor values, and the addition of new layers to the network. The network, after the addition of a new layer, includes links between nodes which skip the hidden layer. The error value used in the backpropagation is reduced from that actually calculated, by adjusting the desired output value, in order to reduce the growth of the weighting factors. After the training of the network, data corresponding to an average of the graphical display of a portion of the shot record, including multiple traces over a period of time, is provided to the network. The time of interest of the data is incremented until such time as the network indicates that the time of interest equals the first break time. The analysis may be repeated for all of the traces in the shot record.
49 Citations
24 Claims
-
1. A method of operating an adaptive network to set values of a plurality of weighting factors therein, comprising:
-
storing data representing a sample input and a desired output in memory; retrieving a portion of said stored data representing said sample input; operating the adaptive network, using the data retrieved in said retrieving step representing the sample input as a data input to the adaptive network, to produce an actual output; calculating a sample error, by determining the difference between the actual output of said adaptive network and the desired output; comparing the sample error to a tolerance limit; responsive to the result of the comparing step indicating that the sample error exceeds the tolerance limit, calculating a learning factor according to the value of the sample error, so that said learning factor increases as the sample error decreases; adjusting at least one of the plurality of weighting factors in said adaptive network by backpropagation according to the generalized delta rule, using the sample error and the learning factor; repeating the retrieving and presenting steps, and the step of calculating the sample error, until said comparing step indicates that the sample error does not exceed the tolerance limit. - View Dependent Claims (2, 3, 4)
-
-
5. A method of operating an adaptive network to set values of a plurality of weighting factors therein, comprising:
-
storing data representing a sample input and a desired output in memory; retrieving a portion of said stored data representing said sample input; operating the adaptive network, using the data retrieved in said retrieving step representing the sample input as a data input to the adaptive network, to produce an actual output; calculating a sample error, by determining the difference between the actual output and the desired output; comparing the sample error to a tolerance limit; responsive to the result of the comparing step indicating that the sample error exceeds the tolerance limit, adjusting at least one of the plurality of weighting factors in said adaptive network according to the sample error; repeating the retrieving, presenting steps, and the step of calculating the sample error a plurality of times; calculating a variance value corresponding to the magnitude of the variation of sample error over a plurality of repetitions of the calculating step; responsive to the variance value being less than a detection limit, and responsive to the sample error exceeding said tolerance limit, modifying said adaptive network; and repeating the retrieving and presenting steps, and the step of calculating the sample error, for the modified adaptive network and the sample input. - View Dependent Claims (6, 7, 8, 9, 10, 11, 12, 13, 14)
-
-
15. A method of operating an adaptive network to set values of a plurality of weighting factors therein, comprising:
-
storing data representing a sample input and a desired output in memory; retrieving a portion of said stored data representing said sample input; operating the adaptive network, using the data retrieved in said retrieving step representing the sample input as a data input to the adaptive network, to produce an actual output; calculating a sample error, by determining the difference between the actual output and the desired output; comparing the sample error to a tolerance limit; responsive to the result of the comparing step indicating that the sample error exceeds the tolerance limit, adjusting at least one of the plurality of weighting factors in said adaptive network by backpropagation according to the generalized delta rule; repeating the retrieving, presenting steps, and the step of calculating the sample error a plurality of times; wherein the value of the desired output used in the adjusting step is modified from the value of the desired output used in calculating the error in such a manner that the error used in the adjusting step is reduced from that calculated.
-
-
16. A method of operating an adaptive network to set values of a plurality of weighting factors therein, comprising:
-
storing data representing a plurality of sample inputs and a plurality of desired outputs, each associated with one of said plurality of sample inputs, in memory; producing a plurality of sample errors, for each of the plurality of sample inputs, by; retrieving a portion of the stored data representing one of the plurality of sample inputs; operating the adaptive network, using the data retrieved in said retrieving step representing the sample input as a data input to the adaptive network, to produce an actual output; calculating a sample error, by determining the difference between the actual output of the adaptive network from the operating step, and its associated desired output; and storing the sample error from the calculating step in memory; wherein, after said step of storing the sample error for the last of the plurality of sample inputs, the plurality of sample errors, each associated with one of the plurality of sample inputs, is stored in memory; comparing the magnitudes of the plurality of sample errors with one another to determine the largest in magnitude of the plurality of sample errors; adjusting at least one of the weighting factors in the adaptive network by backpropagation, using the one of the plurality of sample inputs associated with the largest in magnitude of the plurality of sample errors; and after said adjusting step, repeating said step of producing a plurality of sample errors until the magnitude of the largest in magnitude of the plurality of sample errors is less than a tolerance limit. - View Dependent Claims (17, 18, 19, 20, 21, 22, 23, 24)
-
Specification