learning by fusing heterogeneous data
play

Learning by Fusing Heterogeneous Data Marinka Zitnik Thesis - PowerPoint PPT Presentation

Aug 13, 2022 •124 likes •1.32k views

Learning by Fusing Heterogeneous Data Marinka Zitnik Thesis Defense, October 22 2015 Motivation Marinka Zitnik - PhD Thesis Large Heterogeneous Data Compendia Marinka Zitnik - PhD Thesis Large Heterogeneous Data Compendia Large-scale

  1. A Many Data E B D C A E B Matrices F G Many shared D C factors Solution: DFMF Algorithm Input: A set R of relation matrices R ij ; constraint matrices Θ ( t ) F G for t 2 { 1 , 2 , . . . , max i t i } ; ranks k 1 , k 2 , . . . , k r ( i, j 2 [ r ] ). Output: Matrix factors S and G . 1) Initialize G i for i = 1 , 2 , . . . , r. 2) Repeat until convergence: • Construct R and G using their definitions in Eq. (1) and Eq. (3). • Update S using: S ( G T G ) − 1 G T RG ( G T G ) − 1 . (8) • Set G ( e ) 0 for i = 1 , 2 , . . . , r . i • Set G ( d ) 0 for i = 1 , 2 , . . . , r . i • For R ij 2 R : G ( e ) ij ) + + G i ( S ij G T ( R ij G j S T j G j S T += ij ) − i ij ) − + G i ( S ij G T G ( d ) ( R ij G j S T j G j S T ij ) + += i ij G i S ij ) + + G j ( S T G ( e ) ( R T ij G T += i G i S ij ) − j G ( d ) ij G i S ij ) − + G j ( S T i G i S ij ) + (10) ( R T ij G T += j • For t = 1 , 2 , . . . , max i t i : G ( e ) [ Θ ( t ) ] − G i += for i = 1 , 2 , . . . , r i i G ( d ) [ Θ ( t ) ] + G i += for i = 1 , 2 , . . . , r (11) i i • Construct G as: v v v t G ( e ) t G ( e ) t G ( e ) u u u u r u 1 u 2 G G � Diag( , , . . . , ) , (12) G ( d ) G ( d ) G ( d ) 1 2 r where � denotes the Hadamard product. The p · and · are · entry-wise operations. Marinka Zitnik - PhD Thesis

  2. A Many Data E B D C A E B Matrices F G Many shared D C factors Solution: DFMF Algorithm Input: A set R of relation matrices R ij ; constraint matrices Θ ( t ) F G for t 2 { 1 , 2 , . . . , max i t i } ; ranks k 1 , k 2 , . . . , k r ( i, j 2 [ r ] ). Output: Matrix factors S and G . 1) Initialize G i for i = 1 , 2 , . . . , r. 2) Repeat until convergence: • Construct R and G using their definitions in Eq. (1) and Eq. (3). • Update S using: S ( G T G ) − 1 G T RG ( G T G ) − 1 . (8) • Set G ( e ) 0 for i = 1 , 2 , . . . , r . i • Set G ( d ) 0 for i = 1 , 2 , . . . , r . i • For R ij 2 R : G ( e ) ij ) + + G i ( S ij G T ( R ij G j S T j G j S T += ij ) − i ij ) − + G i ( S ij G T G ( d ) ( R ij G j S T j G j S T ij ) + += i ij G i S ij ) + + G j ( S T G ( e ) ( R T ij G T += i G i S ij ) − j G ( d ) ij G i S ij ) − + G j ( S T i G i S ij ) + (10) ( R T ij G T += j • For t = 1 , 2 , . . . , max i t i : G ( e ) [ Θ ( t ) ] − G i += for i = 1 , 2 , . . . , r i i G ( d ) [ Θ ( t ) ] + G i += for i = 1 , 2 , . . . , r (11) i i • Construct G as: v v v t G ( e ) t G ( e ) t G ( e ) u u u u r u 1 u 2 G G � Diag( , , . . . , ) , (12) G ( d ) G ( d ) G ( d ) 1 2 r where � denotes the Hadamard product. The p · and · are · entry-wise operations. Marinka Zitnik - PhD Thesis

  3. Two Case Studies of Collective Matrix Factorization Marinka Zitnik - PhD Thesis

  4. #1: Amoeba Marinka Zitnik - PhD Thesis

  5. Search for Bacterial Response Genes 50,000 clonal mutants genetic screen Gram+ defective: genome 12,000 genes swp1, gpi, nagB1 found 7 genes Gram- defective: workload 5 years clkB, spc3, alyL, nip7 estimated ~200 genes Nasser et al (2013) Curr Biol Marinka Zitnik - PhD Thesis

  6. Dictyostelium Bacterial Gene Hunt A data-driven approach 14 data sources 4 Gram- seed genes 9 candidate genes Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  7. Dictyostelium Bacterial Gene Hunt 8 ABC family Miranda et al. 2013 7 9 Bacterial Development RNA-seq Parikh et al. 2010 Nasser et al. 2013 A data-driven approach R 1,8 R 1,9 R 1,7 10 14 data sources R 1,10 6 Phenotype 1 4 Gram- seed genes Ontology R 1,6 Reactome 9 candidate genes term Gene pathway Θ 1 R 6,5 R 1,5 R 1,2 R 1,4 5 R 6,4 2 KEGG 3 R 2,3 pathway PubMed MeSH R 5,4 identifier 4 R 2,4 descriptor Gene Ontology term Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  8. Latent Chaining and Profiling 1 Dicty genes G 1 S 1,2 2 Drugs S 2,3 3 Diseases Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  9. Latent Chaining and Profiling 1 Dicty genes G 1 Dicty genes Drugs Diseases S 1,2 2 Drugs S 2,3 3 Diseases Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  10. Latent Chaining and Profiling 1 Dicty genes G 1 Dicty genes Drugs Diseases S 1,2 2 = x x Drugs Profile S 2,3 matrix 3 Diseases Dicty genes Diseases Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  11. 1 Dicty genes Latent Chaining Dicty genes Drugs Diseases G 1 S 1,2 2 = x x Drugs and Profiling S 2,3 Profile matrix 3 Diseases Dicty genes Diseases Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  12. 1 Dicty genes Latent Chaining Dicty genes Drugs Diseases G 1 S 1,2 2 = x x Drugs and Profiling S 2,3 Profile matrix 3 Diseases Dicty genes Diseases 8 ABC family Miranda et al. 2013 7 9 Bacterial Development RNA-seq Parikh et al. 2010 Nasser et al. 2013 R 1,8 R 1,9 R 1,7 10 R 1,10 6 Phenotype 1 Ontology R 1,6 Reactome term Gene pathway Θ 1 R 6,5 R 1,5 R 1,2 R 1,4 5 R 6,4 2 KEGG 3 R 2,3 pathway PubMed MeSH R 5,4 identifier 4 R 2,4 descriptor Gene Ontology term Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  13. 1 Dicty genes Latent Chaining Dicty genes Drugs Diseases G 1 S 1,2 2 = x x Drugs and Profiling S 2,3 Profile matrix 3 Diseases Dicty genes Diseases 8 ABC family Miranda et al. 2013 7 9 Bacterial Development RNA-seq Latent chains Parikh et al. 2010 Nasser et al. 2013 R 1,8 R 1,9 R 1,7 10 R 1,10 6 Phenotype 1 Ontology R 1,6 Reactome term Gene pathway Θ 1 R 6,5 R 1,5 R 1,2 R 1,4 5 R 6,4 2 KEGG 3 R 2,3 pathway PubMed MeSH R 5,4 identifier 4 R 2,4 descriptor Gene Ontology term Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  14. 1 Dicty genes Latent Chaining Dicty genes Drugs Diseases G 1 S 1,2 2 = x x Drugs and Profiling S 2,3 Profile matrix 3 Diseases Dicty genes Diseases Latent chains 8 ABC family Miranda et al. 2013 7 9 Bacterial Development RNA-seq Parikh et al. 2010 Nasser et al. 2013 R 1,8 R 1,9 R 1,7 10 R 1,10 6 Phenotype 1 Ontology R 1,6 Reactome term Gene pathway Θ 1 R 6,5 R 1,5 R 1,2 R 1,4 5 R 6,4 2 KEGG 3 R 2,3 pathway PubMed MeSH R 5,4 identifier 4 R 2,4 descriptor Gene Ontology term Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  15. 1 Dicty genes Latent Chaining Dicty genes Drugs Diseases G 1 S 1,2 2 = x x Drugs and Profiling S 2,3 Profile matrix 3 Diseases Dicty genes Diseases Latent chains 8 ABC family Miranda et al. 2013 7 9 Bacterial Development RNA-seq Parikh et al. 2010 Nasser et al. 2013 R 1,8 R 1,9 R 1,7 10 R 1,10 6 Phenotype 1 Ontology R 1,6 Reactome term Gene pathway Θ 1 R 6,5 R 1,5 R 1,2 R 1,4 5 R 6,4 2 KEGG 3 R 2,3 pathway PubMed MeSH R 5,4 identifier 4 R 2,4 descriptor Gene Ontology term Seed genes i ii Chains Candidate iii gene iv v vi Similarity scoring Similarity score Scored aggregation vii candidate gene viii Seed ix genes Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  16. Dictyostelium Bacterial Gene Hunt Ž itnik et al. PLoS Comp Bio 2015 Marinka Zitnik - PhD Thesis

  17. Dictyostelium Bacterial Gene Hunt cf50-1 sh smlA DDB acbA pt pirA cf rps10 ac abpC sm tirA DDB DDB_G0272184 DDB pikB tr vps46 si pikA rb swp1 DDB ggtA pi DDB_G0288519 DDB pten DG1 DDB_G0288551 ad tra2 DDB DDB_G0286429 DD_ dscA-1 ds cinC gdt udpB pi sfbA DDB modA DDB Ž itnik et al. PLoS Comp Bio 2015 DDB_G0287399 ab Marinka Zitnik - PhD Thesis prmt5

  18. Dictyostelium Bacterial Gene Hunt 10 4 10 3 10 2 10 10 4 10 3 10 2 10 # of D. d cells cf50-1 sh AX4 smlA DDB acbA– acbA pt pirA cf smlA– rps10 ac pikA–/pikB– abpC sm tirA DDB pten– DDB_G0272184 DDB abpC– pikB tr vps46 si modA– pikA rb swp1 DDB cf50-1– ggtA pi tirA– DDB_G0288519 DDB pten DG1 DDB_G0288551 Day 2 Day 3 ad tra2 DDB DDB_G0286429 8/9 predictions correct! DD_ dscA-1 ds 14 data sources cinC gdt udpB 4 Gram- seed genes pi sfbA DDB 9 candidate genes modA DDB Ž itnik et al. PLoS Comp Bio 2015 DDB_G0287399 ab Marinka Zitnik - PhD Thesis prmt5

  19. #2: Functional Genomics Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  20. #2: Functional Genomics 5 4 5 R 45 MeSH PMID Descriptor R 14 R 42 3 1 Experimental R 13 1 Condition 2 Gene 2 R 12 GO Term R 16 R 62 6 6 KEGG Pathway Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  21. #2: Functional Genomics 5 2 4 5 R 45 Pharmacologic MeSH PMID Action Descriptor R 14 R 12 R 42 3 Θ 1 3 1 PMID 1 Experimental R 13 R 13 1 Condition 2 Chemical Gene R 14 2 4 6 R 12 R 46 Depositor Depositor GO Term R 15 Category R 16 R 62 6 5 6 Substructure KEGG Fingerprint Pathway Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  22. #2: Functional Genomics 5 2 4 5 R 45 Pharmacologic MeSH PMID Action Descriptor R 14 R 12 R 42 3 Θ 1 3 1 PMID 1 Experimental R 13 R 13 1 Condition 2 Chemical Gene R 14 2 4 6 R 12 R 46 Depositor Depositor GO Term R 15 Category R 16 R 62 6 5 6 Substructure KEGG Fingerprint Pathway DFMF Prediction task AUC F 1 100 D. discoideum genes 0.799 0.801 1000 D. discoideum genes 0.826 0.823 Whole D. discoideum genome 0.831 0.849 Pharmacologic actions 0.663 0.834 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  23. #2: Functional Genomics 5 2 4 5 R 45 Pharmacologic MeSH PMID Action Descriptor R 14 R 12 R 42 3 Θ 1 3 1 PMID 1 Experimental R 13 R 13 1 Condition 2 Chemical Gene R 14 2 4 6 R 12 R 46 Depositor Depositor GO Term R 15 Category R 16 R 62 6 5 6 Substructure KEGG Fingerprint Pathway DFMF MKL Prediction task AUC AUC AUC F 1 F 1 100 D. discoideum genes 0.799 0.801 0.801 0.781 0.788 1000 D. discoideum genes 0.826 0.823 0.823 0.787 0.798 Whole D. discoideum genome 0.831 0.849 0.849 0.800 0.821 Pharmacologic actions 0.663 0.834 0.834 0.639 0.811 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  24. #2: Functional Genomics 5 2 4 5 R 45 Pharmacologic MeSH PMID Action Descriptor R 14 R 12 R 42 3 Θ 1 3 1 PMID 1 Experimental R 13 R 13 1 Condition 2 Chemical Gene R 14 2 4 6 R 12 R 46 Depositor Depositor GO Term R 15 Category R 16 R 62 6 5 6 Substructure KEGG Fingerprint Pathway DFMF MKL RF Prediction task AUC AUC AUC AUC AUC F 1 F 1 F 1 100 D. discoideum genes 0.799 0.801 0.801 0.781 0.788 0.788 0.761 0.785 1000 D. discoideum genes 0.826 0.823 0.823 0.787 0.798 0.798 0.767 0.788 Whole D. discoideum genome 0.831 0.849 0.849 0.800 0.821 0.821 0.782 0.801 Pharmacologic actions 0.663 0.834 0.834 0.639 0.811 0.811 0.643 0.819 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  25. #2: Functional Genomics 5 2 4 5 R 45 Pharmacologic MeSH PMID Action Descriptor R 14 R 12 R 42 3 Θ 1 3 1 PMID 1 Experimental R 13 R 13 1 Condition 2 Chemical Gene R 14 2 4 6 R 12 R 46 Depositor Depositor GO Term R 15 Category R 16 R 62 6 5 6 Substructure KEGG Fingerprint Pathway DFMF MKL RF tri-SPMF Prediction task AUC AUC AUC AUC AUC AUC AUC F 1 F 1 F 1 F 1 100 D. discoideum genes 0.799 0.801 0.801 0.781 0.788 0.788 0.761 0.785 0.785 0.731 0.724 1000 D. discoideum genes 0.826 0.823 0.823 0.787 0.798 0.798 0.767 0.788 0.788 0.756 0.741 Whole D. discoideum genome 0.831 0.849 0.849 0.800 0.821 0.821 0.782 0.801 0.801 0.778 0.787 Pharmacologic actions 0.663 0.834 0.834 0.639 0.811 0.811 0.643 0.819 0.819 0.641 0.810 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  26. #2: Functional Genomics 5 2 4 5 Pharmacologic R 45 MeSH Action PMID Descriptor R 12 R 14 3 Θ 1 R 42 3 1 PMID 1 R 13 Experimental R 13 1 Condition Chemical 2 R 14 Gene 4 6 2 R R 46 Depositor 12 Depositor R 15 GO Term Category R 16 R 62 6 5 6 Substructure Fingerprint KEGG Pathway DFMF MKL RF tri-SPMF Prediction task AUC AUC AUC AUC AUC AUC AUC F 1 F 1 F 1 F 1 100 D. discoideum genes 0.799 0.801 0.801 0.781 0.788 0.788 0.761 0.785 0.785 0.731 0.724 1000 D. discoideum genes 0.826 0.823 0.823 0.787 0.798 0.798 0.767 0.788 0.788 0.756 0.741 Whole D. discoideum genome 0.831 0.849 0.849 0.800 0.821 0.821 0.782 0.801 0.801 0.778 0.787 Pharmacologic actions 0.663 0.834 0.834 0.639 0.811 0.811 0.643 0.819 0.819 0.641 0.810 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  27. #2: Functional Genomics 5 2 4 5 Pharmacologic R 45 MeSH Action PMID Descriptor R 12 R 14 3 Θ 1 R 42 3 1 PMID 1 R 13 Experimental R 13 1 Condition Chemical 2 R 14 Gene 4 6 2 R R 46 Depositor 12 Depositor R 15 GO Term Category R 16 R 62 6 5 6 Substructure Fingerprint KEGG Pathway DFMF MKL RF tri-SPMF Prediction task AUC AUC AUC AUC AUC AUC AUC F 1 F 1 F 1 F 1 100 D. discoideum genes 0.799 0.801 0.801 0.781 0.788 0.788 0.761 0.785 0.785 0.731 0.724 1000 D. discoideum genes 0.826 0.823 0.823 0.787 0.798 0.798 0.767 0.788 0.788 0.756 0.741 Whole D. discoideum genome 0.831 0.849 0.849 0.800 0.821 0.821 0.782 0.801 0.801 0.778 0.787 Pharmacologic actions 0.663 0.834 0.834 0.639 0.811 0.811 0.643 0.819 0.819 0.641 0.810 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  28. #2: Functional Genomics Mining disease associations 5 2 4 5 Ž itnik et al Scientific Reports 2013 Pharmacologic R 45 MeSH Action PMID Descriptor R 12 R 14 3 Θ 1 R 42 3 1 PMID 1 R 13 Experimental R 13 1 Condition Chemical 2 R 14 Gene 4 6 2 R R 46 Depositor 12 Depositor R 15 GO Term Category R 16 R 62 6 5 6 Substructure Fingerprint KEGG Pathway DFMF MKL RF tri-SPMF Prediction task AUC AUC AUC AUC AUC AUC AUC F 1 F 1 F 1 F 1 100 D. discoideum genes 0.799 0.801 0.801 0.781 0.788 0.788 0.761 0.785 0.785 0.731 0.724 1000 D. discoideum genes 0.826 0.823 0.823 0.787 0.798 0.798 0.767 0.788 0.788 0.756 0.741 Whole D. discoideum genome 0.831 0.849 0.849 0.800 0.821 0.821 0.782 0.801 0.801 0.778 0.787 Pharmacologic actions 0.663 0.834 0.834 0.639 0.811 0.811 0.643 0.819 0.819 0.641 0.810 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  29. #2: Functional Genomics Mining disease associations 5 2 4 5 Ž itnik et al Scientific Reports 2013 Pharmacologic R 45 MeSH Action PMID Descriptor Predicting drug toxicity R 12 R 14 3 Θ 1 R 42 3 1 PMID 1 R 13 Experimental Ž itnik & Zupan Systems Biomedicine 2014 (CAMDA Award) R 13 1 Condition Chemical 2 R 14 Gene 4 6 2 R R 46 Depositor 12 Depositor R 15 GO Term Category R 16 R 62 6 5 6 Substructure Fingerprint KEGG Pathway DFMF MKL RF tri-SPMF Prediction task AUC AUC AUC AUC AUC AUC AUC F 1 F 1 F 1 F 1 100 D. discoideum genes 0.799 0.801 0.801 0.781 0.788 0.788 0.761 0.785 0.785 0.731 0.724 1000 D. discoideum genes 0.826 0.823 0.823 0.787 0.798 0.798 0.767 0.788 0.788 0.756 0.741 Whole D. discoideum genome 0.831 0.849 0.849 0.800 0.821 0.821 0.782 0.801 0.801 0.778 0.787 Pharmacologic actions 0.663 0.834 0.834 0.639 0.811 0.811 0.643 0.819 0.819 0.641 0.810 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  30. #2: Functional Genomics Mining disease associations 5 2 4 5 Ž itnik et al Scientific Reports 2013 Pharmacologic R 45 MeSH Action PMID Descriptor Predicting drug toxicity R 12 R 14 3 Θ 1 R 42 3 1 PMID 1 R 13 Experimental Ž itnik & Zupan Systems Biomedicine 2014 (CAMDA Award) R 13 1 Condition Chemical 2 R 14 Gene 4 6 2 R R 46 Depositor 12 Depositor Predicting gene functions R 15 GO Term Category R 16 R 62 6 5 Ž itnik & Zupan In PSB 2014 6 Substructure Fingerprint KEGG Pathway DFMF MKL RF tri-SPMF Prediction task AUC AUC AUC AUC AUC AUC AUC F 1 F 1 F 1 F 1 100 D. discoideum genes 0.799 0.801 0.801 0.781 0.788 0.788 0.761 0.785 0.785 0.731 0.724 1000 D. discoideum genes 0.826 0.823 0.823 0.787 0.798 0.798 0.767 0.788 0.788 0.756 0.741 Whole D. discoideum genome 0.831 0.849 0.849 0.800 0.821 0.821 0.782 0.801 0.801 0.778 0.787 Pharmacologic actions 0.663 0.834 0.834 0.639 0.811 0.811 0.643 0.819 0.819 0.641 0.810 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  31. #2: Functional Genomics Mining disease associations 5 2 4 5 Ž itnik et al Scientific Reports 2013 Pharmacologic R 45 MeSH Action PMID Descriptor Predicting drug toxicity R 12 R 14 3 Θ 1 R 42 3 1 PMID 1 R 13 Experimental Ž itnik & Zupan Systems Biomedicine 2014 (CAMDA Award) R 13 1 Condition Chemical 2 R 14 Gene 4 6 2 R R 46 Depositor 12 Depositor Predicting gene functions R 15 GO Term Category R 16 R 62 6 5 Ž itnik & Zupan In PSB 2014 6 Substructure Fingerprint KEGG Pathway Predicting cancer survival Ž itnik & Zupan Systems Biomedicine 2015 (CAMDA Award) DFMF MKL RF tri-SPMF Prediction task AUC AUC AUC AUC AUC AUC AUC F 1 F 1 F 1 F 1 100 D. discoideum genes 0.799 0.801 0.801 0.781 0.788 0.788 0.761 0.785 0.785 0.731 0.724 1000 D. discoideum genes 0.826 0.823 0.823 0.787 0.798 0.798 0.767 0.788 0.788 0.756 0.741 Whole D. discoideum genome 0.831 0.849 0.849 0.800 0.821 0.821 0.782 0.801 0.801 0.778 0.787 Pharmacologic actions 0.663 0.834 0.834 0.639 0.811 0.811 0.643 0.819 0.819 0.641 0.810 Ž itnik & Zupan IEEE TPAMI 2015 Marinka Zitnik - PhD Thesis

  32. Key Idea: Transfer of Knowledge Model parameters Marinka Zitnik - PhD Thesis

  33. Key Idea: Transfer of Knowledge Model parameters Marinka Zitnik - PhD Thesis

  34. Key Idea: Transfer of Knowledge Model parameters Objects of one type Data view Marinka Zitnik - PhD Thesis

  35. Key Idea: Transfer of Knowledge Model parameters Objects of one type Data view Marinka Zitnik - PhD Thesis

  36. Key Idea: Transfer of Knowledge Model parameters Objects of one type Data view Marinka Zitnik - PhD Thesis

  37. Key Idea: Transfer of Knowledge Model parameters Objects of one type Data view Marinka Zitnik - PhD Thesis

  38. Key Idea: Transfer of Knowledge Heterogeneous data domain space Model parameters Objects of one type Data view Marinka Zitnik - PhD Thesis

  39. Key Idea: Transfer of Knowledge Heterogeneous data domain space Model parameters Objects of one type Data view Marinka Zitnik - PhD Thesis

  40. Key Idea: Transfer of Knowledge Heterogeneous data domain space Context jumping in the latent space Model parameters Objects of one type Data view Marinka Zitnik - PhD Thesis

  41. Transfer of Knowledge: Another Example Network Inference from Mixed Data Marinka Zitnik - PhD Thesis

  42. Marinka Zitnik - PhD Thesis

  43. Direct inference threshold value Marinka Zitnik - PhD Thesis

  44. Direct inference threshold value Model-based inference model parameters Marinka Zitnik - PhD Thesis

  45. Mixed Data Marinka Zitnik - PhD Thesis

  46. Mixed Data RNA-seq count data count transcripts mapped to genomic locations Marinka Zitnik - PhD Thesis

  47. Mixed Data RNA-seq count data l a i n m o o s s n i i o b P e v i t a g count transcripts e mapped to genomic N locations Marinka Zitnik - PhD Thesis

  48. Mixed Data Somatic mutations RNA-seq count data l a i n m o o s s n i i o b P e v i t a g count transcripts e No mutation mapped to genomic N locations Short indel Single base substitution Marinka Zitnik - PhD Thesis

  49. Mixed Data Somatic mutations l a RNA-seq count data l i a i m n m o o o s s n n i i o b i P t e l v u i t M a g count transcripts e No mutation mapped to genomic N locations Short indel Single base substitution Marinka Zitnik - PhD Thesis

  50. Network Inference from Mixed Data Marinka Zitnik - PhD Thesis

  51. Network Inference from Mixed Data is an object of interest Marinka Zitnik - PhD Thesis

  52. Network Inference from Mixed Data is an object of interest Nodes Edges Marinka Zitnik - PhD Thesis

  53. Network Inference from Mixed Data is an object of interest Nodes Edges Object weights Marinka Zitnik - PhD Thesis

  54. Network Inference from Mixed Data is an object of interest Nodes Edges Object weights Object-object interactions Marinka Zitnik - PhD Thesis

  55. Network Inference from Mixed Data Objective function Marinka Zitnik - PhD Thesis

  56. Network Inference from Mixed Data Objective function Data following distribution Marinka Zitnik - PhD Thesis

  57. Network Inference from Mixed Data Objective function Data following Data following distribution distribution Marinka Zitnik - PhD Thesis

  58. Network Inference from Mixed Data Objective function Latent factor reuse Data following Data following distribution distribution Marinka Zitnik - PhD Thesis

  59. Network Inference from Mixed Data Marinka Zitnik - PhD Thesis

  60. Network Inference from Mixed Data Data Data Marinka Zitnik - PhD Thesis

  61. Network Inference from Mixed Data Data Data Marinka Zitnik - PhD Thesis

  62. Network Inference from Mixed Data Data Data Marinka Zitnik - PhD Thesis

Recommend

Kernel Methods for Fusing Heterogeneous Data Gunnar R atsch Friedrich Miescher Laboratory, Max

Kernel Methods for Fusing Heterogeneous Data Gunnar R atsch Friedrich Miescher Laboratory, Max

SVMs Kernels & the Trick Non-vectorial Data Data Integration Software References Kernel Methods for Fusing Heterogeneous Data Gunnar R atsch Friedrich Miescher Laboratory, Max Planck Society T ubingen, Germany

1.05k views • 81 slides
Fusing Non Fusing Non- -Volumetric, Spatially Volumetric, Spatially- - Localized Data with

Fusing Non Fusing Non- -Volumetric, Spatially Volumetric, Spatially- - Localized Data with

Fusing Non Fusing Non- -Volumetric, Spatially Volumetric, Spatially- - Localized Data with Localized Data with V l V l Volumetric Data Volumetric Data i D i D Robert Weersink, Harsimran Braisch, Greg Bootsma David Jaffray Greg

308 views • 20 slides
Fusing space-time data under measurement error for computer model output Veronica J. Berrocal (

Fusing space-time data under measurement error for computer model output Veronica J. Berrocal (

Fusing space-time data under measurement error for computer model output Veronica J. Berrocal ( vjb2@stat.duke.edu ) SAMSI joint work with Alan E. Gelfand and David M. Holland Veronica J. Berrocal Fusing space-time data under measurement error

577 views • 46 slides
Next Generation Data Discovery Fusing Structured and Unstructured Content from Multiple

Next Generation Data Discovery Fusing Structured and Unstructured Content from Multiple

Next Generation Data Discovery Fusing Structured and Unstructured Content from Multiple Repositories Chris Meredith UDOT Dan Quinn PTFS Questions Shared Drives Vision: Enterprise-wide Data Discovery Information Transparency

763 views • 45 slides
Towards Better Crash Frequency Modeling: Fusing Machine Learning & Econometric Methods

Towards Better Crash Frequency Modeling: Fusing Machine Learning & Econometric Methods

Towards Better Crash Frequency Modeling: Fusing Machine Learning & Econometric Methods Presenter: Behram Wali Ph.D. Student TSITE 2017 Summer Meeting Morning Session July 26, 2017 Contents Background/Challenges Conceptual

693 views • 44 slides
Fusing Continuous and Probabilistic Approach Discrete Data, on the Resulting Location The

Fusing Continuous and Probabilistic Approach Discrete Data, on the Resulting Location The

Case Study Computational . . . What We Plan to Do Available Data: What . . . Fusing Continuous and Probabilistic Approach Discrete Data, on the Resulting Location The Results of the . . . Example of Merging Seismic It May Be Useful to . .

713 views • 17 slides
Fusing point and areal level space-time data with application to wet deposition Alan Gelfand

Fusing point and areal level space-time data with application to wet deposition Alan Gelfand

Fusing point and areal level space-time data with application to wet deposition Alan Gelfand Duke University Joint work with Sujit Sahu and David Holland Alan E. Gelfand Fusing point and areal level space-time data Chemical Deposition

1.4k views • 84 slides
Innovative Special Project of National Significance (SPNS): Fusing Part A, B, C, & D Data for

Innovative Special Project of National Significance (SPNS): Fusing Part A, B, C, & D Data for

Innovative Special Project of National Significance (SPNS): Fusing Part A, B, C, & D Data for MyCareContinuum Dashboard and Empowering Consumers with an Award-Winning Low-Health- Literacy Patient Portal Milagros Izquierdo, Division Director,

1.08k views • 64 slides
Minimax-Angle Learning for Optimal Treatment Decision with Heterogeneous Data Chengchun Shi

Minimax-Angle Learning for Optimal Treatment Decision with Heterogeneous Data Chengchun Shi

Minimax-Angle Learning for Optimal Treatment Decision with Heterogeneous Data Chengchun Shi Department of Statistics North Carolina State University Joint work with Wenbin Lu and Rui Song August 3, 2016 Chengchun Shi (NCSU) Minimax-Angle

907 views • 66 slides
Advanced Approaches to Object Recognition and 3D Model Construction from Heterogeneous Data

Advanced Approaches to Object Recognition and 3D Model Construction from Heterogeneous Data

Advanced Approaches to Object Recognition and 3D Model Construction from Heterogeneous Data Evgeny Burnaev Skoltech, ADASE group Joint with Alexander Notchenko Supervised Deep Learning data Type Supervision 2D Image classification, Class

836 views • 45 slides
Linguistic and Cultural Diversity Reinvented (LINCDIRE): Fusing Indigenous and Western approaches

Linguistic and Cultural Diversity Reinvented (LINCDIRE): Fusing Indigenous and Western approaches

Linguistic and Cultural Diversity Reinvented (LINCDIRE): Fusing Indigenous and Western approaches for plurilingual and pluricultural learning Alan Corbiere (M'Chigeeng First Nation) Sara Potkonjak (York University) on behalf of Enrica Piccardo

335 views • 18 slides
Dual Similarity Learning for Heterogeneous One-Class Collaborative Filtering Xiancong Chen , Weike

Dual Similarity Learning for Heterogeneous One-Class Collaborative Filtering Xiancong Chen , Weike

Dual Similarity Learning for Heterogeneous One-Class Collaborative Filtering Xiancong Chen , Weike Pan, Zhong Ming National Engineering Laboratory for Big Data System Computing Technology, Guangdong Laboratory of Artificial Intelligence and

503 views • 21 slides
Champagne: Champagne: Data Change Propagation for Data Change Propagation for Heterogeneous

Champagne: Champagne: Data Change Propagation for Data Change Propagation for Heterogeneous

Champagne: Champagne: Data Change Propagation for Data Change Propagation for Heterogeneous Information Systems Heterogeneous Information Systems Ralf Rantzau Carmen Constantinescu Uwe Heinkel Holger Meinecke University of Stuttgart

568 views • 7 slides
Normalization of Phenotypic Data from a Clinical Data Warehouse: Case Study of Heterogeneous

Normalization of Phenotypic Data from a Clinical Data Warehouse: Case Study of Heterogeneous

Normalization of Phenotypic Data from a Clinical Data Warehouse: Case Study of Heterogeneous Blood Type Data with Surprising Results James J. Cimino, MD Formerly: Chief of the Laboratory for Informatics Development NIH Clinical Center,

483 views • 29 slides
CoMerge Toward Efficient Data Placement in Shared Heterogeneous Memory Systems Thaleia Dimitra

CoMerge Toward Efficient Data Placement in Shared Heterogeneous Memory Systems Thaleia Dimitra

CoMerge Toward Efficient Data Placement in Shared Heterogeneous Memory Systems Thaleia Dimitra Doudali Ada Gavrilovska Motivation Performance slowdown in heterogeneous memory systems. Application data objects How to reduce the

343 views • 12 slides
Modeling Highly Heterogeneous (Large) Data Sets: Towards a Billion Models Robert Grossman

Modeling Highly Heterogeneous (Large) Data Sets: Towards a Billion Models Robert Grossman

Modeling Highly Heterogeneous (Large) Data Sets: Towards a Billion Models Robert Grossman University of Illinois at Chicago & Open Data Group Traditional Statistics One small data set A few attributes Vector-valued data Data

453 views • 18 slides
Fault-Tolerant Data Collection in Fault-Tolerant Data Collection in Heterogeneous Intelligent

Fault-Tolerant Data Collection in Fault-Tolerant Data Collection in Heterogeneous Intelligent

Fault-Tolerant Data Collection in Fault-Tolerant Data Collection in Heterogeneous Intelligent Monitoring Networks Networks Heterogeneous Intelligent Monitoring Jing Deng Department of Computer Science University of North Carolina at

335 views • 18 slides
Fusing Generic Objectness and Visual Saliency for Salient Object Detection Yasin KAVAK

Fusing Generic Objectness and Visual Saliency for Salient Object Detection Yasin KAVAK

Fusing Generic Objectness and Visual Saliency for Salient Object Detection Yasin KAVAK 06/12/2012 Citation 1: Salient Object Detection: A Benchmark Fusing for Salient Object Detection INDEX (Related Work) [3] B. Alexe, T. Deselaers, and V.

410 views • 29 slides
on Astrophysical Data Processing Heterogeneous Many-Core Systems Theodore Kisner, LBNL

on Astrophysical Data Processing Heterogeneous Many-Core Systems Theodore Kisner, LBNL

on Astrophysical Data Processing Heterogeneous Many-Core Systems Theodore Kisner, LBNL Thursday, December 16, 2010 Astrophysical Data Processing Data Acquisition Data Manipulation / Calculations Data Indexing / Storage Data Selection /

953 views • 11 slides
Bala lancing Efficiency and Fair irness in in Heterogeneous GPU Clu lusters for Deep Learning

Bala lancing Efficiency and Fair irness in in Heterogeneous GPU Clu lusters for Deep Learning

Bala lancing Efficiency and Fair irness in in Heterogeneous GPU Clu lusters for Deep Learning Shubham Chaudhary | Ramachandran Ramjee | Muthian Sivathanu Nipun Kwatra | Srinidhi Viswanatha Microsoft Research India Scheduling of Deep Learning

169 views • 16 slides
Coverage in Heterogeneous Coverage in Heterogeneous Networks Xiaoli Chu King s College

Coverage in Heterogeneous Coverage in Heterogeneous Networks Xiaoli Chu King s College

Coverage in Heterogeneous Coverage in Heterogeneous Networks Xiaoli Chu King s College London UC4G Beijing Workshop August 2010 Outline Introduction Heterogeneous networks Heterogeneous networks Challenges Coverage in

344 views • 32 slides
Pluto A Distributed Heterogeneous Deep Learning Framework Jun Yang, Yan Chen Large Scale

Pluto A Distributed Heterogeneous Deep Learning Framework Jun Yang, Yan Chen Large Scale

Pluto A Distributed Heterogeneous Deep Learning Framework Jun Yang, Yan Chen Large Scale Learning, Alibaba Cloud Outline PAI(Platform of Artificial Intelligence) PAI Overview Deep Learning with PAI Pluto PAI DL Application

655 views • 27 slides
Steganalysis in high dimensions: Fusing classifiers built on random subspaces Jan Kodovsk,

Steganalysis in high dimensions: Fusing classifiers built on random subspaces Jan Kodovsk,

Steganalysis in high dimensions: Fusing classifiers built on random subspaces Jan Kodovsk, Jessica Fridrich January 25, 2011 / SPIE Steganalysis in high dimensions:, Fusing classifiers built on random subspaces 1 / 14 Motivation Modern

536 views • 19 slides
A Massively Scalable Architecture for Learning Representations from Heterogeneous Graphs NVIDIA

A Massively Scalable Architecture for Learning Representations from Heterogeneous Graphs NVIDIA

A Massively Scalable Architecture for Learning Representations from Heterogeneous Graphs NVIDIA GPU Technology Conference 2019 - San Jose, CA C. Bayan Bruss Anish Khazane 1. Overview & Background TODAYS TALK 2. Our Approach How to

680 views • 54 slides

More recommend