We’re rewarding the question askers & reputations are being recalculated! Read more.
    Tweeted twitter.com/StackSciComp/status/670239910872395779
2 added precision tag
source | link

Double precision-precision calculations are significantly /slowerslower or more expensive/ then regular float than single-precision calculations, for. For example nvidia tesla, the NVidia Tesla which performs well on doubles is much more expensive then regular GPU. In

At the same time I do not know Machine Learningabout machine learning cases whenwhere double-precision is really needed. Currently I think that 64bit float's is64-bit floats are needed by the model only in case then it is overfitted inover-fitted. In other words doubles are needed only by "bad" models with bad generalization.

Double precision calculations are significantly /slower or more expensive/ then regular float calculations, for example nvidia tesla which performs well on doubles is much more expensive then regular GPU. In the same time I do not know Machine Learning cases when double is really needed. Currently I think that 64bit float's is needed by model only in case then it is overfitted in other words doubles needed only by "bad" models with bad generalization.

Double-precision calculations are significantly slower or more expensive than single-precision calculations. For example, the NVidia Tesla which performs well on doubles is much more expensive then regular GPU.

At the same time I do not know about machine learning cases where double-precision is really needed. Currently I think that 64-bit floats are needed by the model only in case then it is over-fitted. In other words doubles are needed only by "bad" models with bad generalization.

1
source | link

Are there tasks in machine learning which require double precision floating points?

Double precision calculations are significantly /slower or more expensive/ then regular float calculations, for example nvidia tesla which performs well on doubles is much more expensive then regular GPU. In the same time I do not know Machine Learning cases when double is really needed. Currently I think that 64bit float's is needed by model only in case then it is overfitted in other words doubles needed only by "bad" models with bad generalization.