International Business Machines Corporation

Rensselaer Polytechnic Institute

Tejaswini Pedapati

Pin-Yu Chen

Ruchir Puri

Chuang Gan

John Martin Prager

Srinivasan Parthasarathy

Amit Dhurandhar

Karthikeyan Shanmugam

Bo Wu

Jennifer J Liang

Avinash Balakrishnan

Chunheng Jiang

Jianxi Gao

An output layer is removed from a pre-trained neural network model and a neural capacitance probe unit with multiple layers is incorporated on top of one or more bottom layers of the pre-trained neural network model. The neural capacitance probe unit is randomly initialized and a modified neural network model is trained by fine-tuning the one or more bottom layers on a target dataset for a maximum number of epochs, the modified neural network model comprising the neural capacitance probe unit incorporated with multiple layers on top of the one or more bottom layers of the pre-trained neural network model. An adjacency matrix is obtained from the initialized neural capacitance probe unit and a neural capacitance metric is computed using the adjacency matrix. An active model is selected using the neural capacitance metric and a machine learning system is configured using the active model.

Neural capacitance: neural network selection via edge dynamics

A method, system, and computer program product, including generating a contrastive explanation for a decision of a classifier trained on structured data, highlighting an important feature that justifies the decision, and determining a minimal set of new values for features that alter the decision.

Model agnostic contrastive explanations for structured data

Computerized interactive feature visualization is carried out on a data set&#x2014;a plurality of insight classes rank a plurality of features of the data set. Via a computerized user interface, user feedback is obtained based on the interactive feature visualization&#x2014;a user selects and ranks a subset of the features. At least one transformation function is applied to at least one feature of the subset of features selected by the user, to automatically construct, with a computer, at least one additional feature for the data set. The data set with the at least one additional feature is a transformed data set. In some cases, a supervised task is carried out on the final data set; accuracy of a machine learning system implementing the at least one supervised task can be enhanced by the at least one additional feature, and/or a physical system can be controlled based on results of the at least one supervised task.

Mixed initiative feature engineering

A mechanism is provided in a data processing system comprising at least one processor and at least one memory comprising instructions, which are executed by the at least one processor and configure the processor to implement a patient information extractor. The patient information extractor receives a query specification for executing a query on a patient electronic medical record (EMR). The query specification provides parameters indicating a methodology for extracting search results from the patient EMR. The patient information extractor retrieves the patient EMR from a patient registry. The patient information extractor automatically executes the query specification on the retrieved patient EMR to thereby extract the search results from the patient EMR in accordance with the parameters of the query specification. The patient information extractor automatically processes the extracted search results to generate a patient indicator value. The patient indicator value represents an answer to a question about the patient. A patient evaluation operation is performed based on the patient indicator value.

Extracting patient information from an electronic medical record

Growing community of inventors

New York, NY, United States of America

Tejaswini Pedapati