Less Is More: Adaptive Trainable Gradient Dropout for Deep Neural Networks.

The undeniable computational power of artificial neural networks has granted the scientific community the ability to exploit the available data in ways previously inconceivable. However, deep neural networks require an overwhelming quantity of data in order to interpret the underlying connections between them, and therefore, be able to complete the specific task that they have been assigned to. Feeding a deep neural network with vast amounts of data usually ensures efficiency, but may, however, harm the network's ability to generalize. To tackle this, numerous regularization techniques have been proposed, with dropout being one of the most dominant. This paper proposes a selective gradient dropout method, which, instead of relying on dropping random weights, learns to freeze the training process of specific connections, thereby increasing the overall network's sparsity in an adaptive manner, by driving it to utilize more salient weights. The experimental results show that the produced sparse network outperforms the baseline on numerous image classification datasets, and additionally, the yielded results occurred after significantly less training epochs.

Eighteen years' experience applying old and current strategies in the pre-participation cardiovascular screening of athletes.

INTRODUCTION: Cardiovascular pre-participation screening (PPS) is recommended for the identification of athletes at risk for sudden cardiac death. However, there is currently no universally accepted screening protocol. METHODS: Two distinct PPS strategies were studied in a large cohort of Greek athletes (5 to 39 years old): PPS I, with routine 12-lead ECG and echo, in addition to personal and family history, and physical examination; and PPS II, without routine echo. PPS I (12,353 athletes) was performed from 1992 to 2002, and PPS II (9852 athletes) from 2003 to 2010. RESULTS: "Abnormal" findings were observed in 49.3% of the athletes (49.6% in PPS I and 48.9% in PPS II, p=0.299). Most of them were age- or exercise-related. Further evaluation was recommended for 8.3% of the athletes. Finally, 39 athletes (22 from PPS I) were excluded from competitive sports. Hypertrophic cardiomyopathy was found in 7 athletes. Other abnormalities were: dilated cardiomyopathy; complete heart block; coronary artery disease; Wolf-Parkinson-White syndrome; and severe hypertension. The ECG played a critical role in the exclusion of 13 athletes, compared to only one for echo. Both PPS methods revealed an almost equal incidence of findings. CONCLUSIONS: We suggest that the routine use of ECG alone is sufficient for the successful screening of athletes.