Machine learning for non-imaging data using task-dependent feature dependencies

1 . A computer-implemented method, comprising:

determining, with computer hardware, task-dependent feature dependencies for a group of features extracted from non-imaging data received with the computer hardware;

generating, with the computer hardware, reformatted data by reformatting the non-imaging data based on the task-dependent feature dependencies, wherein

the reformatting converts strongly dependent features among the group of features into one or more subgroups based on task-specific, feature dependency-based priors;

generating, by an image generator, a two-dimensional (2D) image representation of the non-imaging data by configuring the reformatted data into the 2D image representation;

inputting the 2D image representation to a machine learning model implemented with the computer hardware, wherein the machine learning model is trained using the strongly dependent features, of the group of features, that exclude features unrelated to a machine learning task;

generating, using the machine learning model, a machine learning prediction based on the reformatted data as configured in the 2D image representation; and

outputting, by the computer hardware, the machine learning prediction.

2 . The computer-implemented method of claim 1 , further comprising:

configuring the reformatted data into the 2D image representation, wherein

the configuring of the reformatted data comprises positioning the task-specific, feature dependency-based priors belonging to a same subgroup of the one or more subgroups as spatial priors in the 2D image representation, and

each spatial prior of the spatial priors is positioned within a predetermined distance from remaining spatial priors of the spatial priors.

3 . The computer-implemented method of claim 1 , wherein the task-dependent feature dependencies are determined based on pairwise correlations among the group of features.

4 . The computer-implemented method of claim 3 , wherein

the pairwise correlations are conditional correlations, and

the pairwise correlations are conditioned to one class of a plurality of classes associated with the non-imaging data.

5 . The computer-implemented method of claim 1 , wherein the machine learning prediction is performed using a convolutional neural network.

6 . The computer-implemented method of claim 1 , further comprising:

fusing the non-imaging data with imaging data, wherein the imaging data is selected based on a predetermined commonality with the non-imaging data.

7 . The computer-implemented method of claim 6 , wherein

the imaging data is selected from the group consisting of radiological imaging data and pathological imaging data, and

the non-imaging data selected from the group consisting of molecular data, genomic data, clinical data, and demographic data.

8 . A system, comprising:

a processor configured to execute operations comprising:

determining task-dependent feature dependencies for a group of features extracted from non-imaging data received with computer hardware;

generating reformatted data by reformatting the non-imaging data based on the task-dependent feature dependencies, wherein

the reformatting converts strongly dependent features among the group of features into one or more subgroups based on task-specific, feature dependency-based priors;

generating, by an image generator, a two-dimensional (2D) image representation of the non-imaging data by configuring the reformatted data into the 2D image representation;

inputting the 2D image representation to a machine learning model implemented with the computer hardware, wherein the machine learning model is trained using the strongly dependent features, of the group of features, that exclude features unrelated to a machine learning task;

generating, using the machine learning model, a machine learning prediction based on the reformatted data as configured in the 2D image representation; and

outputting, by the computer hardware, the machine learning prediction.

9 . The system of claim 8 , wherein the processor is further configured to execute the operations including

configuring the reformatted data into the 2D image representation, wherein

the configuring of the reformatted data comprises positioning the task-specific, feature dependency-based priors belonging to a same subgroup of the one or more subgroups as spatial priors in the 2D image representation, and

each spatial prior of the spatial priors is positioned within a predetermined distance from remaining spatial priors of the spatial priors.

10 . The system of claim 8 , wherein the task-dependent feature dependencies are determined based on pairwise correlations among the group of features.

11 . The system of claim 10 , wherein

the pairwise correlations are conditional correlations, and

the pairwise correlations are conditioned to one class of a plurality of classes associated with the non-imaging data.

12 . The system of claim 8 , wherein the machine learning prediction is performed using a convolutional neural network.

13 . The system of claim 8 , wherein the processor is further configured to execute the operations including

fusing the non-imaging data with imaging data, wherein the imaging data is selected based on a predetermined commonality with the non-imaging data.

14 . A computer program product, comprising:

one or more computer-readable storage media and program instructions collectively stored on the one or more computer-readable storage media, the program instructions executable by a processor to cause the processor to execute operations comprising:

determining task-dependent feature dependencies for a group of features extracted from non-imaging data received with computer hardware;

generating reformatted data by reformatting the non-imaging data based on the task-dependent feature dependencies, wherein

the reformatting converts strongly dependent features among the group of features into one or more subgroups based on task-specific, feature dependency-based priors;

generating, by an image generator, a two-dimensional (2D) image representation of the non-imaging data by configuring the reformatted data into the 2D image representation;

inputting the 2D image representation to a machine learning model implemented with the computer hardware, wherein the machine learning model is trained using the strongly dependent features, of the group of features, that exclude features unrelated to a machine learning task;

generating, using the machine learning model, a machine learning prediction based on the reformatted data as configured in the 2D image representation; and

outputting, by the computer hardware, the machine learning prediction.

15 . The computer program product of claim 14 , wherein the program instructions further cause the processor to execute the operations including

configuring the reformatted data into the 2D image representation, wherein

the configuring of the reformatted data comprises positioning the task-specific, feature dependency-based priors belonging to a same subgroup of the one or more subgroups as spatial priors in the 2D image representation, and

each spatial prior of the spatial priors is positioned within a predetermined distance from remaining spatial priors of the spatial priors.

16 . The computer program product of claim 14 , wherein the task-dependent feature dependencies are determined based on pairwise correlations among the group of features.

17 . The computer program product of claim 16 , wherein

the pairwise correlations are conditional correlations, and

the pairwise correlations are conditioned to one class of a plurality of classes associated with the non-imaging data.

18 . The computer program product of claim 14 , wherein the machine learning prediction is performed using a convolutional neural network.

19 . The computer program product of claim 14 , wherein the program instructions further cause the processor to execute the operations including

fusing the non-imaging data with imaging data, wherein the imaging data is selected based on a predetermined commonality with the non-imaging data.

20 . The computer program product of claim 19 , wherein

the imaging data is selected from the group consisting of radiological imaging data and pathological imaging data, and

the non-imaging data selected from the group consisting of molecular data, genomic data, clinical data, and demographic data.