Opening book details…
Can I read 2020 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) on EtoBox?
2020 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) by Li, Wei (author);Zhao, Qingqing (author);Zhang, Han (author);Quan, Xiongwen (author);Xu, Jing (author);Yin, Yanbin (author) is a nonfiction available to read on EtoBox.
What is 2020 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) about?
Convolutional neural network (CNN) has been successfully used for the identification of motif occupancy. However, the CNN architecture requires varying length instead of fixed-length filters due to different motif lengths. Moreover, plain neural networks with single point estimation for weights suffer from over-fitting, which is more likely to occur as increasing parameters for multi-scale modeling. Hence, we have designed a Bayesian Multi-scale CNN. The model employs convolutional filters of different scales to extract latent features of DNA sequence, and incorporates Bayesian architecture which regards multi-scale weights as random variables. We further stack two sequential convolutional operations for mean and variance respectively, and apply Bayes by Backprop for posterior estimation of weights. Results have shown that our method not only improved the prediction performance for motif occupancy identification, but also prevented over-fitting due to the capability of Bayesian neural network. The model has also developed a measure of uncertainty estimation for model assessment.
Who reads 2020 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)?
It is typically read by self-directed learners exploring a subject in depth.
Common subject areas: history, science, philosophy, social sciences.
- Author
- Li, Wei (author);Zhao, Qingqing (author);Zhang, Han (author);Quan, Xiongwen (author);Xu, Jing (author);Yin, Yanbin (author)
- Publisher
- IEEE
- Published
- 2020
- Language
- EN
- Category
- nonfiction