Upgrade to Pro — share decks privately, control downloads, hide ads and more …

OpenTalks.AI - Александр Артеменков, Towards Scalable Dimensionality Reduction via Noise Contrastive Estimation

Ad8ae7af280edaecb09bd73a551b5e5f?s=47 OpenTalks.AI
February 21, 2020
Science
0
1.8k

OpenTalks.AI - Александр Артеменков, Towards Scalable Dimensionality Reduction via Noise Contrastive Estimation

Ad8ae7af280edaecb09bd73a551b5e5f?s=128

OpenTalks.AI

February 21, 2020
Tweet

More Decks by OpenTalks.AI

See All by OpenTalks.AI
Ad8ae7af280edaecb09bd73a551b5e5f?s=48 opentalks
0
320
Ad8ae7af280edaecb09bd73a551b5e5f?s=48 opentalks
0
320
Ad8ae7af280edaecb09bd73a551b5e5f?s=48 opentalks
0
310
Ad8ae7af280edaecb09bd73a551b5e5f?s=48 opentalks
0
330
Ad8ae7af280edaecb09bd73a551b5e5f?s=48 opentalks
0
37
Ad8ae7af280edaecb09bd73a551b5e5f?s=48 opentalks
0
300
Ad8ae7af280edaecb09bd73a551b5e5f?s=48 opentalks
0
290
Ad8ae7af280edaecb09bd73a551b5e5f?s=48 opentalks
0
300
Ad8ae7af280edaecb09bd73a551b5e5f?s=48 opentalks
0
290

Other Decks in Science

See All in Science
1896b034326abb4bc429d6718dc6a42e?s=48 shomaekawa
0
220
931bf827a9410634647b1b71f863555c?s=48 highlow
0
680
A2cac4b3dcb2bc0b87917ddc034ef708?s=48 sansandsoc
0
480
638dd29baf6aa478d8eb0aeb0075c404?s=48 sshimizu2006
0
140
A60224e356e52f631bec70bc9df52889?s=48 empitsu
0
450
C612ab39597a17ba5948cae54d13f99f?s=48 mkimura
2
1k
213dacb7e1fa4f74ba845360c2e45c7c?s=48 shel
0
460
58660723de21de826f67924c8498336c?s=48 yasslab
PRO
0
480
6402413ab839ccca6cce08c8bbd51827?s=48 terensarnic
0
240
204f36383109212baaedfabb8abcfc9e?s=48 konakalab
1
330
33ccb53e73a47b600b433992e1e2836f?s=48 agritof80
0
330
8f34aed4a8e8a054f3f5339f047ac1e6?s=48 imamachin
0
150

Featured

See All Featured
433acaea1012b25d97ae66da9b998dad?s=48 malarkey
194
8.8k
A60068bce2e73de3a37ca9d2dbe36092?s=48 bryan
35
3.7k
016edace78ffe826f2348d1e0c504afd?s=48 maggiecrowley
15
750
15f2b8221f247985a27a627d2ece72f0?s=48 pauljervisheath
195
15k
2bb923c57a1fdc3a2eb484bb8d565fd2?s=48 ufuk
60
7.1k
Dad095ea7038f89f760419ce475d5d14?s=48 michaelherold
227
8.8k
Be9caeb9d4ef9944d151af909063ed6e?s=48 samlambert
237
10k
282d599b414022eba5096510f675b6e2?s=48 chrislema
174
14k
A0517b08532aec8ab2aae3f65d40ad86?s=48 hannesfritz
32
1.1k
E190023b66e2b8aa73a842b106920c93?s=48 zenorocha
300
40k
766b9746b59e6f09e280cd33cf4ed419?s=48 skipperchong
13
920
D09fad3e36ae6841f54820be0e907f7a?s=48 rocio
156
12k

Transcript

  1. Towards Scalable Dimensionality Reduction Aleksandr Artemenkov joint work with Maxim

    Panov via Noise Contrastive Estimation

  2. News Headlines Of India 3 million topics you are unfamiliar

    with Kaggle Dataset
  3. None

  4. Just repeat 3 million times! (not a good idea)

  5. Is there a better approach? (visualization)

  6. None

  7. Noise Contrastive Estimation (without a single formula) (check out the

    poster for them)

  8. Building Representation “cold” “frosty” “cold” “warm” “cold” “first” similar different

    unrelated (noisy) ? ? ?

  9. Simple Insight • similar points = closer than predefined noisy

    distance • different points = further than predefined noisy distance Any benefits from such approach?

  10. Parallelized Batch Optimization • noisy distance is predefined • repel

    different points • attract similar points • adjustments are independent of each other to distinguish them from noise Meet NCVis! (Noise Contrastive Visualization) GitHub Repository

  11. Why use NCVis? (there are reasons) Paper

  12. None
  13. None

  14. Thank you for your attention! (find me near the poster)