Latent Tree Models & Statistical Foundation for TCM Nevin L. Zhang Joint Work with: Chen Tao, Wang Yi, Yuan Shihong Department of Computer Science & Engineering The Hong Kong University of Science & Technology

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 2 Publications N. L. Zhang, S. H. Yuan, T. Chen and Y. Wang (2008). Latent tree models and diagnosis in traditional Chinese medicine. Artificial Intelligence in Medicine. 42, N. L. Zhang, S. H. Yuan, T. Chen and Y. Wang (2008). Statistical Validation of TCM Theories. Journal of Alternative and Complementary Medicine. Accepted. N. L. Zhang, S. H. Yuan, T. Chen, and Y. Wang (2007). Hierarchical Latent Class Models and Statistical Foundation for Traditional Chinese Medicine 11th Conference on Artificial Intelligence in Medicine (AIME 07), 07-11, July 2007, Amsterdam, The Netherlands.

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 3 Latent Tree Models (LTM) l Bayesian networks with n Rooted tree structure n Discrete random variables n Leaves observed (manifest variables) n Internal nodes latent (latent variables) l Also known as hierarchical latent class (HLC) models, HLC models P(Y1), P(Y2|Y1), P(X1|Y2), P(X2|Y2), …

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 4 Example l Manifest variables n Math Grade, Science Grade, Literature Grade, History Grade l Latent variables n Analytic Skill, Literal Skill, Intelligence

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 5 Learning Latent Tree Models: The problem X1X2…X6X7 10…11 11…00 01…01 …………… Determine l Number of latent variables l Cardinality of each latent variable l Model Structure l Conditional probability distributions

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 6 Learning Latent Tree Models: The Algorithms l Model Selection n Several scores examined: BIC, BICe, CS, AIC, holdout likelihood n BIC: best choice for the time being l Model optimization n Double hill climbing (DHC), 2002  7 manifest variables. n Single hill climbing (SHC), 2004  12 manifest variables n Heuristic SHC (HSHC), 2004  50 manifest variables n EAST, 2008  As efficient as HSHC, and finds better models

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 7 Traditional Chinese Medicine (TCM) l TCM statement: n Yang deficiency ( 阳虚 ): intolerance to cold ( 畏寒 ), cold limbs ( 肢冷 ), cold lumbus and back ( 腰背冷 ), and so on …. n Regarded by many as not scientific, even groundless. l Two aspects to the meaning 1. Claim: There exists a class of patients, who characteristically have the cold symptoms. The cold symptoms co-occur in a group of people, 2. Explanation offered: Due to deficiency of Yang. It fails to warm the body l What to do? n Previous work focused on 2. n New idea: Do data analysis for 1

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 8 Objectivity of the Claimed Pattern l TCM Claim: there exits a class of patients, in whom symptoms such as ‘intolerance to cold’, ‘cold limbs’, ‘cold lumbus and back’, and so on co-occur at the same time l How to prove or disapprove that such claimed TCM classes exist in the world? n Systematically collect data about symptoms of patients. n Perform cluster analysis, obtain natural clusters of patients n If the natural clusters corresponds to the TCM classes, then YES. 1.Existence of TCM classes validated 2.Descriptions of TCM classes refined and systematically expanded 3.Establish a statistical foundation for TCM

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 9 Why Latent Tree Models? l TCM uses multiple interrelated latent concepts to explain co-occurrence of symptoms n Yang deficiency ( 肾阳虚 ), Yin deficiency ( 肾阴虚 ):, Essence insufficiency ( 肾 精亏虚 ), … TCM theories are latent structure models in natural language. l Need latent structure models n With multiple interrelated latent variables.. l Latent Tree Models are the simplest such models

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 10 Empirical Results l Can we find the claimed TCM classes using latent tree models? n We collected a data set about kidney deficiency ( 肾虚 ) n 35 symptom variables, 2600 records

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 11 Result of Data Analysis l Y0-Y34: manifest variables from data l X0-X13: latent variables introduced by data analysis l Structure interesting, supports TCM’s theories about various symptoms. (Zhang et al. 2008, AI in Medicine)

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 12 Latent Clusters l X1: n 5 states: s0, s1, s2, s3, s4 n Samples grouped into 5 clusters l Cluster X1=s4 {sample | P(X1=s4|sample) > 0.95}  Cold symptoms co-occur in samples l Class implicitly claimed by TCM found! l Description of class refined n By Math vs by words

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 13 Statistical Validation of TCM Theory Experiences TCM Theory Ancient Times Data LT Model

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 14 Other TCM Data Sets l From Beijing U of TCM, 973 project n Depression Depression n Hepatitis B Hepatitis B n Chronic Renal Failure Chronic Renal Failure n … l China Academy of TCM n Subhealth Subhealth n Type 2 Diabetes Type 2 Diabetes l In all cases, claimed TCM classes n Validated n Quantified and refined

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 15 Another Perspective l Just now: validation of TCM theory. l Another perspective: improve diagnosis n TCM diagnosis: classification n Problems: boundaries between classes not clear n Our work is helpful in clarifying the boundaries

Learning Latent Tree Models & TCM ASEAN-China IBW: Page 16 Conclusions l Latent tree models, and latent structure models in general, offer framework for n Density estimation n Latent structure discovery n Multidimensional clustering. n Can play a fundamental role in modernizing TCM n Can be useful in many other areas  Probabilistic inference, classification, semi-supervised learning…  marketing, survey studies, …. l We have only scratched the surface.

Thank You!