Optimizing the Capacity of Extreme Learning Machines for Biomedical Informatics Applications
Optimizing the Capacity of Extreme Learning Machines for Biomedical Informatics Applications
复制标题
优化生物医学信息学应用的极限学习机的能力
DOI：
10.1109/icercs57948.2023.10434204
复制
发表时间：
2023
期刊：
影响因子：
--
通讯作者：
Abhijit Das
中科院分区：
文献类型：
--
作者：
J. Logeshwaran;Rajat Bhardwaj;Shailaja Salagrama;Abhijit Das
本文讨论了一种提高面向生物医学信息学程序的严重深度学习机(ELM)能力的方法。该技术包括改变训练集的维度、隐含层神经元的种类、使用的激活函数的类型、使用的不同类型的神经元的范围以及使用的正则化技术。为了最大限度地发挥ELM方式在生物医学信息学应用中的能力，作者主张使用逐步搜索技术来确定参数的精细组合。这项技术需要首先定义ELM参数的基线或初始配置，然后通过深入学习每个优化步骤的经验效果来迭代地增强该配置。在每个优化步骤都采用了独特的正则化策略，这可能涉及修剪隐藏的神经元，以此来降低模型的复杂性。作者在公开的HAD数据集上验证了他们的技术，这些数据集包括密歇根缺失值推定(MIMV)数据集和来自卫生与公众服务部麻醉学分支的临床记录数据集。
the paper discusses a way of growing the capability of severe deep learning machines (ELM) for biomedical informatics programs. This technique involves varying the dimensions of the training set, the variety of neurons in the hidden layer, the kind of activation function used, the range of different sorts of neurons used and the regularization techniques used.so that it will maximize the ability of ELM fashions for biomedical informatics applications, the authors advocate using a stepwise seek technique to determine the fine combination of parameters. This technique entails first defining a baseline or initial configuration of ELM parameters and then iteratively enhancing that configuration via deep learning from the empirical effects of each optimization step. distinctive regularization strategies are carried out at every optimization step which could involve pruning hidden neurons as a way to reduce the complexity of the model. The authors verified their technique on publicly to be had datasets consisting of the Michigan Imputation of missing Values (MIMV) dataset and on a dataset of clinical records from the branch of Anesthesiology, branch of health and Human services.