u
play

U Berlin Institute of Technology Department Machine Learning - PowerPoint PPT Presentation

Mar 10, 2024 •304 likes •1.22k views

Canonical Trend Analysis for Social Networks Felix Biemann, Jens-Michalis Papaioannou, Mikio Braun, Matthias L. Jugel, Klaus-Robert Mller, Andreas Harth U Berlin Institute of Technology Department Machine Learning Trends Canonical

  1. Canonical Trend Analysis for Social Networks Felix Bießmann, Jens-Michalis Papaioannou, Mikio Braun, Matthias L. Jugel, Klaus-Robert Müller, Andreas Harth U Berlin Institute of Technology Department Machine Learning

  2. Trends Canonical Temporal Dynamics of Web Data 2 U

  3. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) U

  4. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information U

  5. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs U

  6. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs Understanding dynamic graphs [Leskovec et al, KDD, 2005] U

  7. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs Understanding dynamic graphs [Leskovec et al, KDD, 2005] Causal Inference [Lozano and Sindhwani, NIPS 2010] U

  8. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs Understanding dynamic graphs [Leskovec et al, KDD, 2005] Causal Inference [Lozano and Sindhwani, NIPS 2010] Diffusion of information [Gomez Rodriguez et al, ICML 2011/2012] U

  9. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs Understanding dynamic graphs [Leskovec et al, KDD, 2005] Causal Inference [Lozano and Sindhwani, NIPS 2010] Diffusion of information [Gomez Rodriguez et al, ICML 2011/2012] Canonical Trend Analysis U

  10. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs Understanding dynamic graphs [Leskovec et al, KDD, 2005] Causal Inference [Lozano and Sindhwani, NIPS 2010] Diffusion of information [Gomez Rodriguez et al, ICML 2011/2012] Canonical Trend Analysis ‣ Exploits temporal structure to find trends U

  11. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs Understanding dynamic graphs [Leskovec et al, KDD, 2005] Causal Inference [Lozano and Sindhwani, NIPS 2010] Diffusion of information [Gomez Rodriguez et al, ICML 2011/2012] Canonical Trend Analysis ‣ Exploits temporal structure to find trends ‣ Find web sources that precede/follow trends U

  12. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs Understanding dynamic graphs [Leskovec et al, KDD, 2005] Causal Inference [Lozano and Sindhwani, NIPS 2010] Diffusion of information [Gomez Rodriguez et al, ICML 2011/2012] Canonical Trend Analysis Examples: ‣ Exploits temporal structure to find trends ‣ Find web sources that precede/follow trends U

  13. Canonical Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs Understanding dynamic graphs [Leskovec et al, KDD, 2005] Causal Inference [Lozano and Sindhwani, NIPS 2010] Diffusion of information [Gomez Rodriguez et al, ICML 2011/2012] Canonical Trend Analysis Examples: ‣ Exploits temporal structure to find trends ‣ Find web sources that precede/follow trends ‣ Spatiotemporal Dynamics of Retweets to News Articles U

  14. Understanding dynamic graphs [Leskovec et al, KDD, 2005] Trends Temporal Dynamics of Web Data 2 Web content is copied, repeated or rephrased (Trends/Memes) This temporal structure contains important information Growing interest in temporal dynamics of graphs Canonical Causal Inference [Lozano and Sindhwani, NIPS 2010] Diffusion of information [Gomez Rodriguez et al, ICML 2011/2012] Canonical Trend Analysis Examples: ‣ Exploits temporal structure to find trends ‣ Find web sources that precede/follow trends ‣ Spatiotemporal Dynamics of Retweets to News Articles ‣ Music trends on Last.fm U

  15. Canonical Trends Canonical Correlation Analysis 3 Latent Variable (Trend) Features (e.g. Bag of Words, User actions, edge histograms, ...) Z X Y U

  16. Trends Canonical Correlation Analysis 3 Latent Variable (Trend) Canonical Features (e.g. Bag of Words, User actions, edge histograms, ...) w > w > Z y Y x X X Y U

  17. [Jordan 1875], [Hotelling 1936], [Bach and Jordan 2006] Canonical Trends Canonical Correlation Analysis 3 Latent Variable (Trend) Features (e.g. Bag of Words, User actions, edge histograms, ...) w > w > Z y Y x X X Y w > x XY > w y argmax q w > x XX > w x w > y Y Y > w y w x , w y U

  18. Canonical Trends Canonical Trend Model 4 Latent Variable (Trend) Features (e.g. Bag of Words, User actions, edge histograms, ...) faster Z X Y U

  19. Features actions, edge histograms, ...) Trends Canonical Trend Model 4 Latent Variable (Trend) faster Canonical (e.g. Bag of Words, User X w x ( τ ) > X t � τ w > y Y t Z τ X Y U

  20. Trends Canonical An Example on News Trends 5 U

  21. Canonical Trends An Example on News Trends 5 mashable.com slashdot.org techcrunch.com arstechnica.com U

  22. Canonical Trends An Example on News Trends 5 Time t=-1 t=0 mashable.com slashdot.org techcrunch.com arstechnica.com mashable.com slashdot.org techcrunch.com arstechnica.com U

  23. Time Trends An Example on News Trends 5 t=0 t=-1 Canonical X f ∈ R W × T mashable.com slashdot.org techcrunch.com arstechnica.com mashable.com slashdot.org techcrunch.com arstechnica.com U

  24. Time Trends An Example on News Trends 5 t=0 Canonical t=-1 X f ∈ R W × T mashable.com slashdot.org techcrunch.com arstechnica.com mashable.com slashdot.org techcrunch.com X Y = X f 0 f 0 6 = f arstechnica.com U

  25. Predict future content from past content of Trends An Example on News Trends 5 t=0 t=-1 Time of all other web sources Canonical single web source X f ∈ R W × T mashable.com slashdot.org techcrunch.com arstechnica.com mashable.com slashdot.org techcrunch.com X Y = X f 0 f 0 6 = f arstechnica.com U

  26. Time t=0 Trends An Example on News Trends 5 Canonical t=-1 X f ∈ R W × T mashable.com slashdot.org X w x ( τ ) > X t � τ techcrunch.com τ arstechnica.com Z w > y Y t mashable.com slashdot.org techcrunch.com X Y = X f 0 f 0 6 = f arstechnica.com U

  27. Canonical Trends Why Projecting to Canonical Subspace? 6 Easily interpretable: For Text data each canonical direction is a topic [De Bie and Cristianini, 2004] Information theoretic optimal compression [Creutzig 2009] Conversion of canonical correlations to granger causality index [Otter 1991] U

  28. Trends Canonical Canonical Trend Analysis For Social Networks 7 U

  29. Canonical Trends Canonical Trend Analysis For Social Networks 7 Quantifying spatiotemporal retweet response to news content U

  30. Canonical Trends Canonical Trend Analysis For Social Networks 7 Quantifying spatiotemporal retweet response to news content Finding users ahead and following music trends on Last.fm U

  31. Trends Canonical Canonical Trend Analysis For Social Networks 8 U

  32. Canonical Trends Canonical Trend Analysis For Social Networks 8 Some news web site publishes some content ... U

  33. Canonical Trends Canonical Trend Analysis For Social Networks 8 Some news web site publishes some content ... Time ... which is retweeted U t t + τ 1

  34. Canonical Trends Canonical Trend Analysis For Social Networks 8 Some news web site publishes some content ... Time ... which is retweeted ... at different locations U t t + τ 1

  35. Trends Canonical Data Extraction 9 U

  36. Canonical Trends Data Extraction 9 For each news site extract f ∈ { 1 , 2 , . . . , F } U

  37. Canonical Trends Data Extraction 9 For each news site extract Bag-of-Words Features f ∈ { 1 , 2 , . . . , F } X f = [ x f ( t = 1) , . . . , x f ( t = T )] ∈ R W × T U

  38. Canonical Trends Data Extraction 9 For each news site extract Retweet locations Bag-of-Words Features f ∈ { 1 , 2 , . . . , F } X f = [ x f ( t = 1) , . . . , x f ( t = T )] ∈ R W × T Y f = [ y f ( t = 1) , . . . , y f ( t = T )] ∈ R L × T U

  39. Trends Canonical Data Extraction: Retweet Locations 10 U

  40. Canonical Trends Data Extraction: Retweet Locations 10 1. Extract URI of each news article in twitter stream U

Recommend

More recommend