Integrated Gradients
A gradient-based attribution method that assigns feature importance by integrating gradients along a path from a baseline to the input, satisfying a set of axiomatic properties.
Timeline (last 90 days)
Using integrated gradients attribution, the fragment model achieves a hit rate of 1.5 times the random level in recovering known pharmacophores on the BACE target.
Using integrated gradients attribution, the fragment model achieves a hit rate of 3.2 times the random level in recovering known pharmacophores on the PTR1 target.
Integrated Gradients is a gradient-based attribution method proposed by Sundararajan et al. in 2017 that satisfies axioms such as sensitivity and implementation invariance.
In gradient attribution, the number of steps in Integrated Gradients, whether the baseline is all zeros or the mean, whether gradients are smoothed, and how ReLU saturation regions are handled all significantly change the attribution map.
All Facts (4)
Using integrated gradients attribution, the fragment model achieves a hit rate of 1.5 times the random level in recovering known pharmacophores on the BACE target.
50%UnverifiedUsing integrated gradients attribution, the fragment model achieves a hit rate of 3.2 times the random level in recovering known pharmacophores on the PTR1 target.
50%UnverifiedIntegrated Gradients is a gradient-based attribution method proposed by Sundararajan et al. in 2017 that satisfies axioms such as sensitivity and implementation invariance.
50%UnverifiedIn gradient attribution, the number of steps in Integrated Gradients, whether the baseline is all zeros or the mean, whether gradients are smoothed, and how ReLU saturation regions are handled all significantly change the attribution map.
50%