Artificial Intelligence DNA Hypernetworks Biointelligence Lab School of
Artificial Intelligence DNA Hypernetworks Biointelligence Lab School of Computer Sci. & Eng. Seoul National University
From Simple Graphs to Hypergraphs Simple Graph: Bi-connection v 1 v 4 Hypergraph: Multi-connection v 2 v 1 v 3 G = (V, E) V = {v 1, v 2, v 3, …, v 7} E = {E 1, E 2, E 3, E 4, E 5} E 1 = {v 1, v 3, v 4} E 2 = {v 1, v 4} E 3 = {v 2, v 3, v 6} E 4 = {v 3, v 4, v 6, v 7} E 5 = {v 4, v 5, v 7} E 3 E 1 v 2 E 4 E 2 v 3 v 4 v 6 v 5 E 5 v 7 2 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/
From Hypergraphs to Hypernetworks x 1 x 2 x 15 x 3 x 14 x 13 x 5 x 12 x 6 x 11 x 7 • Vertices: • Genes • Proteins • Chemicals • Words • Hyperedges: • Interactions • Genetic • Signaling • Relations • Associations x 10 x 8 x 9 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/ 3
Hypernetworks l l l A hypernetwork is a hypergraph of weighted edges. It is defined as a triple H = (V, E, W), where V = {v 1, v 2, …, vn}, E = {E 1, E 2, …, En}, and W = {w 1, w 2, …, wn}. An m-hypernetwork consists of a set V of vertices and a subset E of V[m], i. e. H = (V, V[m], W) where V[m] is a set of subsets of V whose elements have precisely m members and W is the set of weights associated with the hyperedges. A hypernetwork H is said to be k-uniform if every edge Ei in E has cardinality (or order) k. A hypernetwork H is k-regular if every vertex has degree k. Note: An ordinary graph is a 2 -uniform hypergraph with wi=1. 4 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/
Generating Hyperedges by Random Sampling A Data Sample x 0=0 Hyperedges A hypernetwork x 0=0 x 1=0 y=1 x 1=0 x 5=1 x 0=0 x 3=1 x 2=0 x 4=0 x 7=0 y=1 x 0=0 x 2=1 x 7=0 y=1 x 1=0 x 2=1 x 6=1 y=1 x 3=1 x 4=0 y=1 x 5=1 x 6=1 x 7=0 y=1 label 5 © 2009, SNU CSE Biointelligence Lab, http: //bi. snu. ac. kr/
Basic Elements of Hypernetworks l A hypernetwork can be interpreted as a library of weighted hyperedges l The cardinality (or order) of each hyperedge in the library may be uniform or hybrid l Weights of hyperedges are updated during random matching process based on training instances 6 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/
Properties of the Hypernetwork Model l Hypernetwork learning addresses both ¨ Structural learning (size and contents of hyperedges) and ¨ Parameter learning (weights of hyperedges) l Probabilistic ¨ Higher-order probabilistic relationship ¨ Overcoming the weakness of the ordinary Bayesian networks l Descriptive ¨ Can discover the building blocks ¨ Higher-order description ¨ Self-organizing random graphs l Useful for Discovery Cognitive ¨ Memory chunk-like storage mechanism (hyperedges as chunks) ¨ Associative recall of the memory Useful for Modeling ¨ A random association graph of chunk-like memory fragments 7
Note: Pattern Recognition by DNA Computer [Zhang, DNA-2006] 8
Note: Molecular Self-Assembly of Hypernetworks xi xj y Molecular Encoding Hypernetwork Representation X 1 X 2 X 8 X 3 X 7 X 4 X 6 X 5 DNA 9
Note: Learning the Hypernetwork (by Evolution) Next generation Library of combinatorial molecules Library Example + Select the library elements matching the example Amplify the matched library elements by PCR [Zhang, DNA 11] Hybridize 10
Memory, Learning, and Mining Learning Raw Data Memory Mining Rules 11 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/
Machine Learning and Human Learning l Machine Learning ¨ Recognition ¨ Short-lived ¨ Local or global ¨ Batch learning ¨ Repetition ¨ Statistical ¨ Connectionist l Human Learning ¨ Recall ¨ Long-lasting ¨ Local and global ¨ Incremental ¨ Sequential ¨ Categorical ¨ Compositional 12 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/
DNA hypernetworks as Cognitive Recall Memory
Hypergraph Models of Recall Memory x 1 x 2 x 15 x 3 • Linguistic Memory • Vertex: Word x 14 x 13 • Edge: Semantic link • Visual Memory x 12 • Vertex: Object (Pixel) x 5 • Edge: Spatial link x 6 x 11 x 7 x 10 x 8 x 9 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/ 14
Visual Memory “Mental Chemistry” Classes 0 -9, Random Sampling of Features of Order 5 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/ 15
Visual Memory: Image Completion “Mental Chemistry” Subsampled Features for Two Classes © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/ 16
Linguistic Memory [Park & Zhang, 2007] 17 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/
Linguistic Memory: Sentence Completion “Mental Chemistry” ¨ Completion (Generation) & Classification (Recognition) Examples Query Completion who are you Corpus: Friends, 24, Prison Break ? are you who ? you who are ? what are you who are you need to wear it Corpus: 24, Prison Break, House ? need to wear it you ? to wear it you need ? wear it you need to ? it you need to wear ? i need to wear it you want to wear it you need to do it you need to wear a Classification Friends 24 24 24 House 24 18 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/
Applications of Cognitive Learning l Biomedical Diagnosis and Treatments ¨ Modeling and Treatment of Memory Deficits ¨ Diagnosis of Language Disorder l Human Computer Interaction (HCI) ¨ Cognitively-Friendly User Interface ¨ Tracking User Interests (Games, Web Applications) l Memory and Learning Research ¨ Data Mining for Learning and Memory Research ¨ Modeling Cognitive Behaviors l Linguistic/Visual Data Processing ¨ Learning and Mining from Text ¨ Learning from Pictures 19 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/
Conclusion l The DNA hypernetworks are a useful tool for modeling cognitive learning and memory processes: ¨ Hypergraphs as memory organization ¨ Random graph process as learning l DNA hypernetworks as recognition memories: ¨ Digit recognition ¨ Face recognition ¨ Movie identification l DNA hypernetworks as recall memories: ¨ Visual memory ¨ Linguistic memory l The multimodal memory game is a scalable platform for studying cognitive learning architectures and algorithms. 20 © 2009, SNU Biointelligence Lab, http: //bi. snu. ac. kr/
- Slides: 20