Closed jamesgoulet closed 8 months ago
Replaced the approximate formulation for the mReLU covariance by the exact formulation from Huber (2020, https://arxiv.org/abs/2009.01730).
I have tested the CPU and GPU versions for both the current as well and the new pyTAGI_v1
@lhnguyen102 I realized I need to update the unit tests ... I will do it and reopen the pull request.
Replaced the approximate formulation for the mReLU covariance by the exact formulation from Huber (2020, https://arxiv.org/abs/2009.01730).
I have tested the CPU and GPU versions for both the current as well and the new pyTAGI_v1