Efficient Neural Network Loss Landscape Generation

Posted by r3tex on Monday April 29, 2019 @11:32AM
from the nextjournal.com dept.

( 2 comments | 25 points | on HN )

Re:Efficient Neural Network Loss Landscape Generation by shele on Monday April 29, 2019 @12:28PM (#19779830)

This brings a wide spectrum of the Julia stack together and underlines the inherent interoperability of Julia solutions:

Flux - Elegant Machine Learning Stack

Makie - High level plotting on the GPU

CuArrays - GPU arrays without code changes

WebSockets - Websockets protocol in Julia

WebIO - Simple abstraction for displaying and interacting with content

Nextjournal - Making this reproducible online

[ Parent | on HN ]

Re:Efficient Neural Network Loss Landscape Generation by anshuligh on Monday April 29, 2019 @01:33PM (#19780552)

Cool article!

In Makie, you can efficiently put a label on a slider by using

`sliderplot, observable = AbstractPlotting.textslider(range, label, start = startvalue)`

where sliderplot is the actual Scene object and observable is the value of the observable.

[ Parent | on HN ]