tfp.substrates.numpy.math.log1mexp

Compute log(1 - exp(-|x|)) elementwise in a numerically stable way.

Main aliases

tfp.experimental.substrates.numpy.math.log1mexp

 
tfp.substrates.numpy.math.log1mexp(
    x, name=None
)

x Float Tensor.
name Python str name prefixed to Ops created by this function. Default value: None (i.e., 'log1mexp').

log1mexp Float Tensor of log1mexp(x).

References

[1]: Machler, Martin. Accurately computing log(1 - exp(-|a|)) https://cran.r-project.org/web/packages/Rmpfr/vignettes/log1mexp-note.pdf