Selected article for: "convolutional layer and filter activation"

Author: Jakub M Bartoszewicz; Anja Seidel; Bernhard Y Renard
Title: Interpretable detection of novel human viruses from genome sequencing data
  • Document date: 2020_1_30
  • ID: ac00tai9_39
    Snippet: Once we calculate the contributions of convolutional filters for the first layer, Ï• (yj ) i (z, x) for the first convolutional layer of a network with one-hot encoded inputs and an all-zero reference can be efficiently calculated using weight matrices and filter activation differences (Eq. 3-4). First, in this case we do not traverse any non-linearities and can directly use the linear rule (Shrikumar et al., 2019a) to calculate the contributions.....
    Document: Once we calculate the contributions of convolutional filters for the first layer, Ï• (yj ) i (z, x) for the first convolutional layer of a network with one-hot encoded inputs and an all-zero reference can be efficiently calculated using weight matrices and filter activation differences (Eq. 3-4). First, in this case we do not traverse any non-linearities and can directly use the linear rule (Shrikumar et al., 2019a) to calculate the contributions of x i to y j as a product of the weight w i and the input x i . Second, the input values may only be 0 or 1.

    Search related documents:
    Co phrase search for related documents
    • Try single phrases listed below for: 1