Control dataset of clinician-adjudicated/-annotated "true negatives" consisting of 300 stable, 10-minute datastream segments without alarm-triggering features [~3,000 datastream minutes over 300 non-alarm segments with EKG …
Derivation dataset with clinician-adjudicated red alarms [~8,530 datastream minutes over 853 adjudicated red alarms; no EKG or PPG waveform annotations]
Derivation data subset with clinician-adjudicated red alarms and clinician-annotated waveforms [~12,340 datastream minutes over 1,234 adjudicated red alarms with EKG and PPG waveform annotations]
Training (exploratory) data subset for potential validation testing of experimental algorithms to be derived from ATOMICS-0/-1/-2 [~2,520 datastream monitor-hours; no red alarm adjudications or EKG …
This study compares the performance of machine learning to conventional logistic regression for risk factor analysis in pediatric supracondylar humerus fractures using a multi-institutional database.
This is an updated version of original (DOI: 10.26300/1pad-7574) that contains minor revisions (see page iv for list of revisions). Reinforcement learning defines the problem …
This thesis examines measures of sample complexity as a means to evaluate the effectiveness and reliability of machine learning algorithms. We give an overview of …