From data to numbers to knowledge: semantic embeddings By Alvaro Barbero
âą
1 likeâą399 views
https://www.bigdataspain.org/2016/program/thu-from-data-numbers-knowledge-semantic-embeddings.html https://www.youtube.com/watch?v=JjCHvKMgkkY&list=PL6O3g23-p8Tr5eqnIIPdBD_8eE5JBDBik&index=30&t=3s
Recommended
More Related Content
Viewers also liked
Viewers also liked (20)
Similar to From data to numbers to knowledge: semantic embeddings By Alvaro Barbero
Similar to From data to numbers to knowledge: semantic embeddings By Alvaro Barbero (6)
More from Big Data Spain
More from Big Data Spain (20)
Recently uploaded
Recently uploaded (20)
From data to numbers to knowledge: semantic embeddings By Alvaro Barbero
- 2. 7 de diciembre de 2016 Data ï Numbers ï Knowledge Ălvaro Barbero â Chief Data Scientist www.iic.uam.es
- 3. www.iic.uam.es 3 How do we see the world?
- 4. www.iic.uam.es 4 Understanding correlations The more, the better Linear correlations
- 5. www.iic.uam.es 5 Understanding correlations âMoral virtue is a meanâ Aristotle Convex correlations
- 6. www.iic.uam.es 6 Understanding correlations Few is bad, but slightly more is very good, unless you skip the sweet spot and then you are doing terribly, but if you keep going you get better until you do worse again and after that nothing changes and you do so-so Nonlinear correlations
- 7. www.iic.uam.es THE REAL WORLD IS NON-LINEAR Shocking revelations Big data scientists rampage the cities after the newly discovered uselessness of their linear models. Lorem ipsum. Ea pro natum invidunt repudiandae, his et facilisis vituperatoribus. Mei eu ubique altera senserit, consul eripuit accusata has ne. Ignota verterem te nam, eu cibo causae menandri vim. Sit rebum erant dolorem et, sed odio error ad.Vel molestie corrumpit deterruisset ad, mollis ceteros ad sea. In libris graecis appetere mea. At vim odio lorem omnes, pri id iuvaret partiendo. Vivendo menandri et sed. Lorem volumus blandit cu has.Sit cu alia porro fuisset. Ea pro natum invidunt repudiandae, his et facilisis vituperatoribus. Mei eu ubique altera senserit, consul eripuit accusata has ne. Ignota verterem te nam, eu cibo causae menandri vim. Sit rebum erant dolorem et, sed odio error ad.Vel molestie corrumpit deterruisset ad, mollis ceteros ad sea. BIG DATA TIMES VERY VERY BIG DATA SCIENCE - Since 1802
- 8. www.iic.uam.es 8 The modelling dilemma
- 9. www.iic.uam.es 9 The modelling dilemma Easy to understand and explain Linear models May not faithfully represent reality Non-linear models Accurate representations Very difficult interpretation
- 11. www.iic.uam.es 11 Brain power Low level observations Multilayered non-linear processing Abstract yet linear concepts
- 12. www.iic.uam.es 12 Abstraction examples << Art quality
- 13. www.iic.uam.es 13 Embeddings: abstract non-linearity away Low level observations High level embedding Artificial multilayered non-linear processing Deep Network
- 15. www.iic.uam.es 15 word2vec cat chills on a mat cat chills mushroom a mat Socher et al (2013)
- 16. www.iic.uam.es 16 Exploiting the linearity: semantic algebra King man woman queen Obama USA Russia Putin human animal ethics paella Spain Italy risotto Cristiano Madrid Barcelona Messi
- 17. www.iic.uam.es 17 Bilingual word2vec Word Space Wzh Wen English Words Mandarin Words Socher et al (2013)
- 18. www.iic.uam.es 18 Text embedding models through Recurrent Neural Networks High dimensional representation of a sequence 0.1 0.5 1.0 0.0 2.4 The lazy brown fox Sutskever et al - Sequence to Sequence Learning with neural networks
- 19. www.iic.uam.es 19 Example: books embeddings Cilibrasi, VitĂĄnyi - Normalized Web Distance and Word Similarity Jonathan Swift Oscar Wilde William Shakespeare
- 20. www.iic.uam.es 20 Embedding artistic styles Gatys et al â A Neural Algorithm of Artistic Style Bottou et al - Optimization Methods for Large-Scale Machine Learning Low level observations Style embedding
- 21. www.iic.uam.es 21 Artistic styles as linear relations Google Research - https://research.googleblog.com/2016/10/supercharging-style-transfer.html
- 22. www.iic.uam.es 22 An application: furniture embedding Bell and Bala - Learning visual similarity for product design with convolutional neural networks
- 24. www.iic.uam.es 24 Take home message Reality is highly non- linear Deep Learning can abstract complexity away Complex relations become easy comparisons!
- 25. www.iic.uam.es
- 26. www.iic.uam.es www.iic.uam.es Ălvaro Barbero JimĂ©nez Chief Data Scientist at Instituto de IngenierĂa del Conocimiento (IIC) Elementos grĂĄficos de apoyo obtenidos en: @albarjip Alvaro Barbero