Reverse-complement parameter sharing improves deep learning models for genomics

70 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Opportunities and obstacles for deep learning in biology and medicine (2017, methodology): “To handle this issue, specialized reverse complement convolutional models share parameters to find motifs in both directions [167].”
Biological Sequence Modeling with Convolutional Kernel Networks (2017, neutral): “Other versions have been then considered, by using a fully connected layer that allows mixing information from the two DNA strands [Shrikumar et al., 2017b], or by considering several hidden layers instead of a single one [Zeng et al., 2016].”
Separable Fully Connected Layers Improve Deep Learning Models For Genomics (2017, methodology): “In genomics, the same positional activation patterns are relevant across multiple convolutional channels, and different neurons in the fully connected layer must learn these patterns independently (Shrikumar et al. [2017b]).”
A guide to machine learning for biologists (2021, background): “If the known structure of a biological system can be exploited and neural networks used to learn the unknown parts, then increasingly heavily parameterised models can be replaced with simpler ones that are more amenable to interpretation and more robust on new data [102].”