A number of authors have indicated in recent years that the course of depression is not as favourable as previously expected. Research conducted in order to identify predictors of recovery has shown widely different results. In this paper a sample of 90 consecutive patients with non-chronic major depressive disorders (index episode < 6 months) attending four mental health centres in Madrid were followed up prospectively for 6 months, and clinical social and cognitive variables were studied. The patients were treated pharmacologically and controlled. The rate of recovery was measured according to the Hamilton Rating Scale for Depression (HAM-D). Other tools used were: Life Events and Chronic Difficulties, Global Assessment Functioning in the 6 months prior to the onset of episode, Brown Rating Scale for Self-Esteem and Mannheim Interview of Social Support. The results showed that 41 cases recovered (HAM-D score < 8), 29 cases achieved a partial remission, and major depressive disorder persisted in 17 cases (HAM-D score > or = 18). The presence of personality disorders, having suffered a previous episode, GAF score and some aspects of social support were the variables most associated with non full remission in the logistic regression analysis. Personality disorders and the initial HAM-D score were related to non-improvement. Some clinical and cognitive variables maintain a weak relation to outcome and are rejected in logistic regression. This study emphasizes the relationship of personality, and social variables such as social support and previous global functioning, with incomplete recovery in major depression.
This paper analyzes the effects of approximate multiplication when performing inferences on deep convolutional neural networks (CNNs). The approximate multiplication can reduce the cost of the underlying circuits so that CNN inferences can be performed more efficiently in hardware accelerators. The study identifies the critical factors in the convolution, fully-connected, and batch normalization layers that allow more accurate CNN predictions despite the errors from approximate multiplication. The same factors also provide an arithmetic explanation of why bfloat16 multiplication performs well on CNNs. The experiments are performed with recognized network architectures to show that the approximate multipliers can produce predictions that are nearly as accurate as the FP32 references, without additional training. For example, the ResNet and Inception-v4 models with Mitch-w6 multiplication produces Top-5 errors that are within 0.2% compared to the FP32 references. A brief cost comparison of Mitch-w6 against bfloat16 is presented where a MAC operation saves up to 80% of energy compared to the bfloat16 arithmetic. The most far-reaching contribution of this paper is the analytical justification that multiplications can be approximated while additions need to be exact in CNN MAC operations.
Deciding on the continuous treatment of chronic diseases is vital in terms of economy, quality of life, and time. We present a holistic data mining approach that addresses the prediction of the therapeutic response in a panoramic and feedback way while unveiling relevant medical factors. Panoramic prediction makes it possible to decide whether the treatment will be beneficial without using previous knowledge and without involving unnecessary treatments. Feedback prediction can be more accurate prediction since it considers the results of previous stages of the treatment. A novel label encoding called simulated annealing and rounding (SAR) encoding is also proposed to help improve the accuracy of prediction in both approaches. To unveil the medical factors that make the treatment effective for patients, various techniques are applied to the prediction models found through the proposed approaches. Finally, this methodology is applied in the realistic scenario of analyzing electronic medical records of migraineurs under BoNT-A treatment. The results show a significant improvement in accuracy due to the use of SAR encoding, from close to 60% (baseline) to 75% with panoramic prediction, and up to around 90% when using feedback prediction. Furthermore, the following factors have been found to be relevant when predicting the migraine treatment responses: migraine time evolution, unilateral pain, analgesic abuse, headache days, and the retroocular component. According to doctors, these factors are also medically relevant and in alignment with the medical literature.INDEX TERMS Multi-target prediction, classification algorithms, data mining, simulated annealing.
The Posit™ Number System was introduced in 2017 as a replacement for floating-point numbers. Since then, the community has explored its application in Neural Network related tasks and produced some unit designs which are still far from being competitive with their floating-point counterparts. This paper proposes a Posit Logarithm-Approximate Multiplication (PLAM) scheme to significantly reduce the complexity of posit multipliers, the most power-hungry units within Deep Neural Network architectures. When comparing with state-of-the-art posit multipliers, experiments show that the proposed technique reduces the area, power, and delay of hardware multipliers up to 72.86%, 81.79%, and 17.01%, respectively, without accuracy degradation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.