Back to results
US 12,755,148 B2Granted
Knowledge expansion for improving machine learning
AssigneeInternational Business Machines Corporation
InventorsDinesh C. Verma, Franck Vinh Le, Michele Merler, Dhiraj Joshi, Supriyo Chakraborty, Seraphin Bernard Calo
Life of this patent
Estimating the term…
Abstract
A system may receive an existing base set of knowledge, train a neural network on the base set of knowledge, deploy the neural network on a new data set, generate, using the deployment, instances of new knowledge, and validate the instances of new knowledge.
Claims16
A system comprising:
- a memory;
- and
- a processor in communication with the memory, the processor being configured to perform a process comprising:
- receiving an existing base set of knowledge that has been structured, classified, and annotated;
- training a neural network on the base set of knowledge by propagating the base set through the neural network, identifying output errors based on comparisons with expected results, and updating weights applied to weighting functions of respective neurons that contributed to the output errors to reduce future contributions to the output errors;
- deploying the neural network by applying the trained network to a non-training data set that has not been previously reviewed or annotated and that is not included in the base set of knowledge;
- filtering, using a filter, the non-training data set to remove superfluous data, wherein the filter is selected from a group consisting of a skip list filter, an invalid value filter, and a threshold filter;
- generating, using the deployment, instances of new knowledge from the non-training data set;
- and
- validating the instances of new knowledge.
7 more elements
A method comprising:
- receiving an existing base set of knowledge that has been structured, classified, and annotated;
- training a neural network on the base set of knowledge by propagating the base set through the neural network, identifying output errors based on comparisons with expected results, and updating weights applied to weighting functions of respective neurons that contributed to the output errors to reduce contributions to the output errors;
- deploying the neural network by applying the trained network to a non-training data set that has not been previously reviewed or annotated and that is not included in the base set of knowledge;
- filtering, using a filter, the non-training data set to remove superfluous data, wherein the filter is selected from a group consisting of a skip list filter, an invalid value filter, and a threshold filter;
- generating, using the deployment, instances of new knowledge from the non-training data set;
- and
- validating the instances of new knowledge.
4 more elements
A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method, the method comprising:
- receiving an existing base set of knowledge that has been structured, classified, and annotated;
- training a neural network on the base set of knowledge by propagating the base set through the neural network, identifying output errors based on comparisons with expected results, and updating weights applied to weighting functions of respective neurons that contributed to the output errors to reduce contributions to the output errors;
- deploying the neural network by applying the trained network to a non-training data set that has not been previously reviewed or annotated and that is not included in the base set of knowledge;
- filtering, using a filter, the non-training data set to remove superfluous data, wherein the filter is selected from a group consisting of a skip list filter, an invalid value filter, and a threshold filter;
- generating, using the deployment, instances of new knowledge from the non-training data set;
- and
- validating the instances of new knowledge.
4 more elements
Dependencies are inferred from the claim text.