Automatic processing of ambiguously labeled data
A mechanism is provided that implements a cognitive data processing system for automatically processing ambiguously labeled data associated with a medical image. The cognitive data processing system receives an ambiguously labeled set of training data in which the ambiguously labeled set of training data comprises portions of data and associated labels, and wherein at least one portion of data in the ambiguously labeled set of training data has a plurality of different labels that together render the portion of data ambiguously labeled. The cognitive data processing system configures an implementation of a model that comprises a loss term, a maximizing term, and a sparsity term. The cognitive data processing system processes the ambiguously labeled set of training data based on the model to identifying a mapping that minimizes a loss function and thereby train the cognitive data processing system.
1. A method, in a data processing system comprising at least one processor and at least one memory, the at least one memory comprising instructions that are executed by the at least one processor to cause the at least one processor to be configured to implement a cognitive data processing system for automatically processing ambiguously labeled data associated with a medical image, the method comprising:
receiving, by the cognitive data processing system, an ambiguously labeled set of training data in which the ambiguously labeled set of training data comprises portions of data and associated labels, and wherein at least one portion of data in the ambiguously labeled set of training data has a plurality of different labels that together render the portion of data ambiguously labeled;
configuring, by the cognitive data processing system, an implementation of a model that comprises a loss term, a maximizing term, and a sparsity term, wherein the loss term operates to evaluate candidate classes of data types in which portions of data in the ambiguously labeled set of training data may be classified, wherein the maximizing term operates to identify a single candidate class for a portion of data having ambiguous labels, and wherein the sparsity term operates to assess a number of non-zero elements for ambiguous labels; and
training the cognitive data processing system, via a deep learning process, to identify abnormalities or anatomical structures in medical images by at least processing, by the cognitive data processing system, the ambiguously labeled set of training data based on the model to identify a mapping that minimizes a loss function.
2. The method of claim 1 , wherein the loss function treats all portions of data as negative contributors to all classes of data types outside of the ambiguous set of labels associated with the data, all portions of data having non-ambiguous labels as positive contributors to a single candidate class of data type, and all portions of data having ambiguous labels as positive contributors to one and only one class of data type inside of the ambiguous set.
3. The method of claim 1 , wherein the plurality of different labels comprises one correct label.
4. The method of claim 1 , further comprising:
performing, by the cognitive data processing system, an inference operation based on the trained cognitive data processing system.
5. The method of claim 4 , wherein the inference operation is a medical image classification operation in which one or more anatomical structures or abnormalities in a medical image are classified.
6. The method of claim 1 , wherein the sparsity term is weighted by value σ.
7. The method of claim 1 , wherein the cognitive data processing system is a neural network and wherein the neural network comprises a category of network topologies, including a multi-class perceptron classifier; a traditional neural networks with one input layer, one hidden layer, and one output layer; and deep neural network.
8. A computer program product comprising a computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a cognitive data processing system, causes the cognitive data processing system to:
receive, by the cognitive data processing system, an ambiguously labeled set of training data in which the ambiguously labeled set of training data comprises portions of data and associated labels, and wherein at least one portion of data in the ambiguously labeled set of training data has a plurality of different labels that together render the portion of data ambiguously labeled;
configure, by the cognitive data processing system, an implementation of a model that comprises a loss term, a maximizing term, and a sparsity term, wherein the loss term operates to evaluate candidate classes of data types in which portions of data in the ambiguously labeled set of training data may be classified, wherein the maximizing term operates to identify a single candidate class for a portion of data having ambiguous labels, and wherein the sparsity term operates to assess a number of non-zero elements for ambiguous labels; and
train the cognitive data processing system, via a deep learning process, to identify abnormalities or anatomical structures in medical images by at least processing, by the cognitive data processing system, the ambiguously labeled set of training data based on the model to identify a mapping that minimizes a loss function.
9. The computer program product of claim 8 , wherein the loss function treats all portions of data as negative contributors to all classes of data types outside of the ambiguous set of labels associated with the data, all portions of data having non-ambiguous labels as positive contributors to a single candidate class of data type, and all portions of data having ambiguous labels as positive contributors to one and only one class of data type inside of the ambiguous set.
10. The computer program product of claim 8 , wherein the plurality of different labels comprises one correct label.
11. The computer program product of claim 8 , wherein the computer readable program further causes the cognitive data processing system to:
perform, by the cognitive data processing system, an inference operation based on the trained cognitive data processing system.
12. The computer program product of claim 11 , wherein the inference operation is a medical image classification operation in which one or more anatomical structures or abnormalities in a medical image are classified.
13. The computer program product of claim 8 , wherein the sparsity term is weighted by value σ.
14. The computer program product of claim 8 , wherein the cognitive data processing system is a neural network and wherein the neural network comprises a category of network topologies, including a multi-class perceptron classifier, a traditional neural networks with one input layer, one hidden layer, and one output layer; and deep neural network.
15. A cognitive data processing system comprising:
a processor, and
a memory coupled to the processor, wherein the memory comprises instructions which, when executed by the processor, cause the processor to:
receive, by the cognitive data processing system, an ambiguously labeled set of training data in which the ambiguously labeled set of training data comprises portions of data and associated labels, and wherein at least one portion of data in the ambiguously labeled set of training data has a plurality of different labels that together render the portion of data ambiguously labeled;
configure, by the cognitive data processing system, an implementation of a model that comprises a loss term, a maximizing term, and a sparsity term, wherein the loss term operates to evaluate candidate classes of data types in which portions of data in the ambiguously labeled set of training data may be classified, wherein the maximizing term operates to identify a single candidate class for a portion of data having ambiguous labels, and wherein the sparsity term operates to assess a number of non-zero elements for ambiguous labels; and
train the cognitive data processing system, via a deep learning process, to identify abnormalities or anatomical structures in medical images by at least processing, by the cognitive data processing system, the ambiguously labeled set of training data based on the model to identify a mapping that minimizes a loss function.
16. The cognitive data processing system of claim 15 , wherein the loss function treats all portions of data as negative contributors to all classes of data types outside of the ambiguous set of labels associated with the data, all portions of data having non-ambiguous labels as positive contributors to a single candidate class of data type, and all portions of data having ambiguous labels as positive contributors to one and only one class of data type inside of the ambiguous set.
17. The cognitive data processing system of claim 15 , wherein the plurality of different labels comprises one correct label.
18. The cognitive data processing system of claim 15 , wherein the instructions further causes the processor to:
perform, by the cognitive data processing system, an inference operation based on the trained cognitive data processing system.
19. The cognitive data processing system of claim 18 , wherein the inference operation is a medical image classification operation in which one or more anatomical structures or abnormalities in a medical image are classified.
20. The cognitive data processing system of claim 15 , wherein the sparsity term is weighted by value σ.