Published on November 15, 2016 by Google

Check out g.co/aiexperiments to learn more.

This experiment helps visualize what’s happening in machine learning. It allows coders to see and explore their high-dimensional data. The goal is to eventually make this an open-source tool within TensorFlow, so that any coder can use these visualization techniques to explore their data. g.co/aiexperiments

Built by Daniel Smilkov, Fernanda Viégas, Martin Wattenberg, and the Big Picture team at Google.

More resources:
www.tensorflow.org

Leave a Reply

33 Comments on "A.I. Experiments: Visualizing High-Dimensional Space"

Notify of
avatar

ti inu
Guest
ti inu
3 days 19 hours ago

Hmm interesting Google and thank you for sharing.

EpicClone5082
Guest
EpicClone5082
5 days 23 hours ago

I bet it can't read my bad handwriting

TheAether
Guest
TheAether
14 days 20 hours ago

This is so. cool. °v° Thank you for making all of this stuff publicly available!

mastersoftoday
Guest
mastersoftoday
15 days 15 hours ago

Hey, can you just hang on with this thing? I translate documents for a living and this is pretty scary. Please stop it, Google!

Fandyus CZ
Guest
Fandyus CZ
18 days 4 hours ago

SkyNet is coming soon, pals…

Mikael Murstam
Guest
Mikael Murstam
19 days 13 hours ago

Nice explanation. It all makes sense why you built a tensor accelerating unit now.

Mr.BrYcE
Guest
Mr.BrYcE
20 days 8 hours ago

what should i be learning/praticing before i can tackle tensorflow ?

Major League Pegasus // Alex
Guest
Major League Pegasus // Alex
21 days 10 hours ago

Hacked Roblox servers…

Thomas Lowry (CopperPen)
Guest
Thomas Lowry (CopperPen)
23 days 6 hours ago

i wonder, could you use T-SNE to partially decode the Voiniech manuscript?

joelface
Guest
joelface
10 days 13 hours ago

It's without a doubt that with T-SNE you'd learn valuable data about the manuscript that people studying the manuscript without such software would not pick up on or be able to so clearly demonstrate. I say that with very little knowledge about the Voiniech manuscript, but I still believe it would work.

Thomas Lowry (CopperPen)
Guest
Thomas Lowry (CopperPen)
15 days 12 hours ago

The Manuscript is 240 pages long. and about a quarter of those pages are entirely text; about a third of the pages are half-text half image and the rest is just images with labels.So i think that would constitute as long enough.

Sobsz
Guest
Sobsz
15 days 15 hours ago

Thomas Lowry If it's long enough, one could try the Zipf law.

Thomas Lowry (CopperPen)
Guest
Thomas Lowry (CopperPen)
19 days 13 hours ago

+Mikael Murstam that is something that is still heavily under discussion. But if the text were to be gibberish then using T-SNE we could tell.

Mikael Murstam
Guest
Mikael Murstam
19 days 13 hours ago

it has more or less been proven to be a very old fake.

Snajperis
Guest
Snajperis
23 days 16 hours ago

I uderstand that it is still being developed and not available for use, right? If so, when it would be possible to try out?

Noname Plsno (VanPanzer)
Guest
Noname Plsno (VanPanzer)
20 days 21 hours ago

no,it is available. Just google tensorflow.

wpDiscuz