- Full text PDF available (8)
- This year (6)
- Last five years (7)
BACKGROUND The number of total body nevi is a major risk marker for malignant melanoma. No previous study has evaluated the accuracy of whole body large nevus (> or = 5 mm) self-counts. OBJECTIVE Our purpose was to evaluate the accuracy of large nevus self-counts by sex, age, educational level, body site, family history of skin cancer, and nevus… (More)
Muscle mass is a determining factor in skeletal muscle function and is affected by inactivity, immobilization, disease, and aging. The aim of this study was to develop an objective and time-efficient method to quantify the volume and cross-sectional area of human calf muscles using three-dimensional magnetic resonance images. We have estimated the errors… (More)
We describe a mechanism for subsampling sequences and show how to compute its expected output so that it can be trained with standard backpropagation. We test this approach on a simple toy problem and discuss its shortcomings.
Machine learning models are often used at test-time subject to constraints and trade-offs not present at training-time. For example, a computer vision model operating on an embedded device may need to perform real-time inference, or a translation model operating on a cell phone may wish to bound its average compute time in order to be power-efficient. In… (More)