2020/07/28 by David H. Ledbetter, Eugene Laksana, Ledbetter, David +5
Computer Science · #Applications (stat.AP) #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning in Healthcare
paper · pdf · doi:10.48550/arxiv.2007.14520
openalex publication_date 2020/07/28 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Predictive models in acute care settings must be able to immediately\nrecognize precipitous changes in a patient's status when presented with data\nreflecting such changes. Recurrent neural networks (RNNs) have become common\nfor training and deploying clinical decision support models. They frequently\nexhibit a delayed response to acute events. New information must propagate\nthrough the RNN's cell state memory before the total impact is reflected in the\nmodel's predictions. This work presents input data perseveration as a method of\ntraining and deploying an RNN model to make its predictions more responsive to\nnewly acquired information: input data is replicated during training and\ndeployment. Each replication of the data input impacts the cell state and\noutput of the RNN, but only the output at the final replication is maintained\nand broadcast as the prediction for evaluation and deployment purposes. When\npresented with data reflecting acute events, a model trained and deployed with\ninput perseveration responds with more pronounced immediate changes in\npredictions and maintains globally robust performance. Such a characteristic is\ncrucial in predictive models for an intensive care unit.\n