Meta-learning model training based on causal transportability between datasets

1 . A method, executed by a processor, comprising:

receiving a plurality of datasets related to a plurality of application domains;

determining feature dependency information associated with a first dataset of the plurality of datasets, based on a first user input;

determining feature difference information associated with the first dataset and a second dataset of the received plurality of datasets, based on a second user input and a set of ethical requirements, the second dataset is different from the first dataset;

determining a set of structural causal models (SCMs) associated with the first dataset based on the feature dependency information and the determined feature difference information;

determining a set of ethical coefficients associated with the set of ethical requirements based on an application of a causal transportability model on the set of SCMs;

determining a trust score associated with the first dataset based on the set of ethical coefficients;

capturing, by the set of SCMs, causal effects associated with the plurality of datasets and the feature dependency information and the feature difference information between various attributes between the plurality of datasets;

training a meta-learning model associated with the plurality of application domains, based on a plurality of trust scores associated with the plurality of datasets, the plurality of trust scores includes the trust score associated with the first dataset without requiring access to one or more datasets.

2 . The method according to claim 1 , wherein the feature dependency information includes at least one of:

a set of input variables associated with the first dataset,

a set of output variables associated with the first dataset,

dependency information associated with the set of input variables and the set of output variables, or

a confidence level associated with the dependency information.

3 . The method according to claim 1 , wherein

the feature difference information corresponds to a population difference between a set of first features associated with the first dataset and a set of second features associated with the second dataset, and

the set of first features corresponds to the set of second features.

4 . The method according to claim 1 , wherein the set of ethical requirements includes at least one of: a set of fairness requirements, a set of privacy requirements, or a set of security requirements.

5 . The method according to claim 1 , wherein

each of the set of SCMs corresponds to a directed acyclic graph representative of variables associated with the first dataset, and

the directed acyclic graph is augmented based on a set of selection variables indicative of the feature difference information.

6 . The method according to claim 5 , wherein a population associated with the first dataset is switchable to a population associated with the second dataset, based on an application of a set of conditions on different values of at least one of the set of selection variables.

7 . The method according to claim 5 , wherein the directed acyclic graph is indicative of a causal effect of a set of input variables associated with the first dataset on a set of output variables associated with the first dataset.

8 . The method according to claim 5 , wherein the directed acyclic graph further includes a set of unobserved variables that affect a set of input variables associated with the first dataset.

9 . The method according to claim 1 , wherein each of the set of ethical coefficients is determined based on a confidence level associated with the feature difference information, and a causal transportability coefficient associated with the first dataset and the second dataset.

10 . The method according to claim 9 , wherein

the causal transportability coefficient is determined by use of the causal transportability model, and

the causal transportability model corresponds to a causal fusion tool.

11 . The method according to claim 1 , wherein the trust score is determined based on a weighted average of the set of ethical coefficients.

12 . The method according to claim 1 , wherein the meta-learning model corresponds to a generative learning model including at least one of: a metric-based learning model, an optimization-based learning model, or a model-based learning model.

13 . One or more non-transitory computer-readable storage media configured to store instructions that, in response to being executed, cause an electronic device to perform operations, the operations comprising:

receiving a plurality of datasets related to a plurality of application domains;

determining feature dependency information associated with a first dataset of the plurality of datasets, based on a first user input;

determining feature difference information associated with the first dataset and a second dataset of the received plurality of datasets, based on a second user input and a set of ethical requirements, the second dataset is different from the first dataset;

determining a set of structural causal models (SCMs) associated with the first dataset based on the feature dependency information and the feature difference information;

determining a set of ethical coefficients associated with the set of ethical requirements based on an application of a causal transportability model on the set of SCMs;

determining a trust score associated with the first dataset based on the set of ethical coefficients;

capturing, by the set of SCMs, causal effects associated with the plurality of datasets and the feature dependency information and the feature difference information between various attributes between the plurality of datasets;

training a meta-learning model associated with the plurality of application domains, based on a plurality of trust scores associated with the plurality of datasets, the plurality of trust scores includes the trust score associated with the first dataset without requiring access to one or more datasets.

14 . The one or more non-transitory computer-readable storage media according to claim 13 , wherein the feature dependency information includes at least one of:

a set of input variables associated with the first dataset,

a set of output variables associated with the first dataset,

dependency information associated with the set of input variables and the set of output variables, or

a confidence level associated with the dependency information.

15 . The one or more non-transitory computer-readable storage media according to claim 13 , wherein

the feature difference information corresponds to a population difference between a set of first features associated with the first dataset and a set of second features associated with the second dataset, and

the set of first features corresponds to the set of second features.

16 . The one or more non-transitory computer-readable storage media according to claim 13 , wherein the set of ethical requirements includes at least one of: a set of fairness requirements, a set of privacy requirements, or a set of security requirements.

17 . The one or more non-transitory computer-readable storage media according to claim 13 , wherein

each of the set of SCMs corresponds to a directed acyclic graph representative of variables associated with the first dataset, and

the directed acyclic graph is augmented based on a set of selection variables indicative of the feature difference information.

18 . The one or more non-transitory computer-readable storage media according to claim 17 , wherein a population associated with the first dataset is switchable to a population associated with the second dataset, based on an application of a set of conditions on different values of at least one of the set of selection variables.

19 . The one or more non-transitory computer-readable storage media according to claim 13 , wherein each of the set of ethical coefficients is determined based on a confidence level associated with the feature difference information, and a causal transportability coefficient associated with the first dataset and the second dataset.

20 . An electronic device, comprising:

a memory storing instructions;

a processor, coupled to the memory, that executes the instructions to perform a process comprising:

receiving a plurality of datasets related to a plurality of application domains;

determining feature dependency information associated with a first dataset of the plurality of datasets, based on a first user input;

determining feature difference information associated with the first dataset and a second dataset of the received plurality of datasets, based on a second user input and a set of ethical requirements, the second dataset is different from the first dataset;

determining a set of structural causal models (SCMs) associated with the first dataset based on the feature dependency information and the feature difference information;

determining a set of ethical coefficients associated with the set of ethical requirements based on an application of a causal transportability model on the set of SCMs;

determining a trust score associated with the first dataset based on the set of ethical coefficients;

capturing, by the set of SCMs, causal effects associated with the plurality of datasets and the feature dependency information and the feature difference information between various attributes between the plurality of datasets;

training a meta-learning model associated with the plurality of application domains, based on a plurality of trust scores associated with the plurality of datasets, the plurality of trust scores includes the trust score associated with the first dataset without requiring access to one or more datasets.