Systems and methods for probabilistic geology inversion with neural generative models
In an embodiment, input data (1) associated with a geological area and (2) including a plurality of types of input data is received. The input data is input into a multi-mode embedding encoder configured to generate an output (1) based on the input data, (2) representing a compressed version of the input data and (3) that is in a format compatible with a generative network. First random noise is generated. The output and the first random noise are input to the generative network to generate a first geological model. Second random noise different than the first random noise is generated. The second random noise, the output, and the first geological model is input to the generative network to generate a second geological model different than the first geological model.
1 . A method, comprising:
receiving input data (1) associated with a geological area and (2) including a plurality of types of input data;
inputting the input data into a multi-mode embedding encoder configured to generate an output (1) based on the input data, (2) representing a compressed version of the input data and (3) that is in a format compatible with a generative network;
generating first random noise;
inputting the output and the first random noise to the generative network to generate a first geological model;
generating second random noise different than the first random noise;
inputting the second random noise, the output, and the first geological model to the generative network to generate a second geological model different than the first geological model;
generating third random noise;
inputting the output and the third random noise to the generative network to generate a third geological model;
generating a fourth random noise;
inputting the fourth random noise, the output, and the third geological model to the generative network to generate a fourth geological model different than the third geological model;
inputting the second geological model and not the first geological model into a surrogate model to generate a first surrogate model output associated with the plurality of types of input data;
comparing the first surrogate model output to the input data to determine a first likelihood that the first surrogate model output will occur;
in response to the first likelihood being within a predetermined acceptable range, including the second geological model in a set of geological models;
in response to the first likelihood not being within the predetermined acceptable range, refraining from including the second geological model in the set of geological models;
inputting the fourth geological model and not the third geological model into the surrogate model to generate a second surrogate model output associated with the plurality of types of input data;
comparing the second surrogate model output to the input data to determine a second likelihood that the second surrogate model output will occur;
in response to the second likelihood being within the predetermined acceptable range, including the fourth geological model in the set of geological models;
in response to the second likelihood not being within the predetermined acceptable range, refraining from including the fourth geological model in the set of geological models, the first geological model and the third geological model not included in the set of geological models; and
selecting, via posterior sampling, at least one geological model from the set of geological models.
2 . The method of claim 1 , wherein inputting the output and the first random noise to the generative network to generate the first geological model includes inputting the output, the first random noise, and a fifth geological model to the generative network to generate the first geological model.
3 . The method of claim 1 , wherein the plurality of types of input data includes at least one of geophysics data associated with the geological area, geospatial data associated with the geological area, drill core data associated with the geological area, geological data associated with the geological area, or a geological map associated with the geological area.
4 . The method of claim 1 , wherein the multi-mode embedding encoder includes a first plurality of networks and a second network, and inputting the input data into the multi-mode embedding encoder includes:
inputting the input data into the first plurality of networks to generate a plurality of intermediate outputs; and
inputting the plurality of intermediate outputs into the second network to generate the output.
5 . The method of claim 4 , wherein each network from the first plurality of networks receives a type of input data from the plurality of types of input data different than remaining networks from the first plurality of networks.
6 . The method of claim 4 , wherein the first plurality of networks includes a first network being a first network type and a second network being a second network type different than the first network type.
7 . The method of claim 1 , further comprising:
deleting the first geological model.
8 . An apparatus, comprising:
a memory; and
a processor operatively coupled to the memory, the processor configured to:
receive input data (1) associated with a geological area and (2) including a plurality of types of input data;
input the input data into an encoder configured to generate an output (1) based on the input data, (2) representing a compressed version of the input data and (3) that is in a format compatible with a generative network;
generate first random noise;
input the output and the first random noise to the generative network to generate a first geological model;
generate second random noise different than the first random noise;
input the second random noise, the output, and the first geological model to the generative network to generate a second geological model different than the first geological model;
generate third random noise;
input the output and the third random noise to the generative network to generate a third geological model;
generate a fourth random noise;
input the fourth random noise, the output, and the third geological model to the generative network to generate a fourth geological model different than the third geological model;
input the second geological model and not the first geological model into a surrogate model to generate a first surrogate model output associated with the plurality of types of input data;
compare the first surrogate model output to the input data to determine a first likelihood that the first surrogate model output will occur;
in response to the first likelihood being within a predetermined acceptable range, include the second geological model in a set of geological models;
in response to the first likelihood not being within the predetermined acceptable range, refrain from including the second geological model in the set of geological models;
input the fourth geological model and not the third geological model into the surrogate model to generate a second surrogate model output associated with the plurality of types of input data;
compare the second surrogate model output to the input data to determine a second likelihood that the second surrogate model output will occur;
in response to the second likelihood being within the predetermined acceptable range, include the fourth geological model in the set of geological models;
in response to the second likelihood not being within the predetermined acceptable range, refrain from including the fourth geological model in the set of geological models, the first geological model and the third geological model not included in the set of geological models; and
select, via posterior sampling, at least one geological model from the set of geological models.
9 . The apparatus of claim 8 , wherein inputting the output and the first random noise to the generative network to generate the first geological model includes inputting the output, the first random noise, and a fifth geological model to the generative network to generate the first geological model.
10 . The apparatus of claim 8 , wherein the plurality of types of input data includes at least one of geophysics data associated with the geological area, geospatial data associated with the geological area, drill core data associated with the geological area, geological data associated with the geological area, or a geological map associated with the geological area.
11 . The apparatus of claim 8 , wherein the encoder includes a first plurality of networks and a second network, and inputting the input data into the encoder includes:
inputting the input data into the first plurality of networks to generate a plurality of intermediate outputs; and
inputting the plurality of intermediate outputs into the second network to generate the output.
12 . The apparatus of claim 11 , wherein each network from the first plurality of networks receives a type of input data from the plurality of types of input data different than remaining networks from the first plurality of networks.
13 . The apparatus of claim 11 , wherein the first plurality of networks includes a first network being a first network type and a second network being a second network type different than the first network type.
14 . A non-transitory, processor-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to:
receive input data (1) associated with a geological area and (2) including a plurality of types of input data;
input the input data into a multi-mode embedding encoder configured to generate an output based on the input data;
generate first random noise;
input the output and the first random noise to a generative network to generate a first geological model;
generate second random noise different than the first random noise;
input the second random noise, the output, and the first geological model to the generative network to generate a second geological model different than the first geological model;
generate third random noise;
input the output and the third random noise to the generative network to generate a third geological model;
generate a fourth random noise;
input the fourth random noise, the output, and the third geological model to the generative network to generate a fourth geological model different than the third geological model;
input the second geological model and not the first geological model into a surrogate model to generate a first surrogate model output associated with the plurality of types of input data;
compare the first surrogate model output to the input data to determine a first likelihood that the first surrogate model output will occur;
in response to the first likelihood being within a predetermined acceptable range, include the second geological model in a set of geological models;
in response to the first likelihood not being within the predetermined acceptable range, refrain from including the second geological model in the set of geological models;
input the fourth geological model and not the third geological model into the surrogate model to generate a second surrogate model output associated with the plurality of types of input data;
compare the second surrogate model output to the input data to determine a second likelihood that the second surrogate model output will occur;
in response to the second likelihood being within the predetermined acceptable range, include the fourth geological model in the set of geological models;
in response to the second likelihood not being within the predetermined acceptable range, refrain from including the fourth geological model in the set of geological models, the first geological model and the third geological model not included in the set of geological models; and
select, via posterior sampling, at least one geological model from the set of geological models.
15 . The non-transitory, processor-readable medium of claim 14 , wherein the plurality of types of input data includes at least one of geophysics data associated with the geological area, geospatial data associated with the geological area, drill core data associated with the geological area, geological data associated with the geological area, or a geological map associated with the geological area.
16 . The non-transitory, processor-readable medium of claim 14 , wherein the multi-mode embedding encoder includes a first plurality of networks and a second network, and the instructions to input the input data into the multi-mode embedding encoder include instructions to:
input the input data into the first plurality of networks to generate a plurality of intermediate outputs; and
input the plurality of intermediate outputs into the second network to generate the output.
17 . The non-transitory, processor-readable medium of claim 16 , wherein each network from the first plurality of networks receives a type of input data from the plurality of types of input data different than remaining networks from the first plurality of networks.
18 . The non-transitory, processor-readable medium of claim 16 , wherein the first plurality of networks includes a first network being a first network type and a second network being a second network type different than the first network type.
19 . The non-transitory, processor-readable medium of claim 14 , wherein the plurality of types of input data includes at least three of geophysics data associated with the geological area, geospatial data associated with the geological area, drill core data associated with the geological area, geological data associated with the geological area, or a geological map associated with the geological area.
20 . The non-transitory, processor-readable medium of claim 14 , wherein the plurality of types of input data includes at least five of geophysics data associated with the geological area, geospatial data associated with the geological area, drill core data associated with the geological area, geological data associated with the geological area, or a geological map associated with the geological area.