#226
  • Created on
    2008-11-12 19:08
  • Reported by
    bergstra
  • Status
    assigned
  • Priority
    Low (4)
  • Assigned to
    pascanur
  • Milestone
    0.6.1
  • Plan Level
    -
  • Due Date
    -

Attachments

Related Tickets

Add new subtask or bug

integrate the xlogx file into tensor ops

Make optimizations that insert this Op for all instances of x* log(x). Also replace y * log(x) with the related op.

Numpy implements numpy.log(0)* 0 == NaN and that's not what we want in this case.

This optimization is required for the tutorials.
User picture
  on 2010-03-03 17:19 *
By james.bergstra
Priority changed from Low (4) to Highest (1)
User picture
  on 2010-03-03 19:17 *
By pascanur
Status changed from New to assigned
User picture
  on 2010-03-29 18:52 *
By pascanur
Priority changed from Highest (1) to Low (4)
We decided that from a mathematical point of you replacing y* log(x) does not make sense (which was actually what we need). x * log(x) makes sense but is rarely needed, so we decided to postpone it for some other time until it becomes evident where we need it.
User picture
  on 2011-03-12 04:18 *
By lamblin
We decided that from a mathematical point of you replacing y* log(x) does not make sense (which was actually what we need). x * log(x) makes sense but is rarely needed, so we decided to postpone it for some other time until it becomes evident where we need it.
User picture
  on 2012-01-12 19:28 *
By nouiz
Milestone changed from 0.1 to 0.5.1