Meeting Notes 20090113 Phone meeting (Dejing, Gwen, Bob and Haishan; 1/13/2009)

From Nemo

Jump to: navigation, search


Phone meeting (Dejing, Gwen, Bob and Haishan)

  1. Review issues with TKDD paper draft (replication of case studies from KDD07)
  2. Discuss mapping work with simulated data (for HBM09 abstract submission)
  3. Review to-do list for next month 

  1. Section 4 of TKDD paper: Data mining case studies (methods & results)
    1. LP1/LP2 study replication - Haishan was able to reproduce the intermediate steps in Jaiwei's analysis and is writing these up in a technical report. The end results are a little different (why?)
    2. WL3a study replication -The 25 attributes of LP1 and LP2 data are basically the subset of 31 attributes of Wl3a data. Haishan run clustering based on 31 attributes Bob pointed out. The result is similar to kdd'07 but the total number of pattern factors are still different. In addition, we have not been able to determine where the numbers in the KDD summary table (which add up to 708) are from. We have checked input data (WL3a_PCAautolabel_2007Feb07v4.xls), as well as numbers reported in the CIN paper, but have not been able to replicate these. Haishan suggested there may have been a bug in Jiawei's script.
  2. PCA/ICA and Metric mapping/cross-lab pattern mapping (for HBM09 abstract & paper) 
    1. The mapping discovery between ICA and PCA and between different sets of metrics can be useful for ontology development. The idea is to design the concepts to cover the heterogeneous data from different methods and labs. Dejing also pointed out the mapping discovery and ontology development can be interactive in a sense that the initial ontology design may be helpful for mapping discovery as well.
    2. We will submit to abstracts for HBM09 - one that reports results from simulated ERP data (PCA/ICA and metric mapping) and one that reports results from mapping across patterns from different experiments/labs 
  3. To-do list for ~Jan 13-Feb 1 
    1. Wednesday (Jan 14), we prepare two abstracts for HBM'09. One is about mapping between PCA and ICA. Another one is about mapping between different groups. Haishan will see whether he can get some concrete results before the submission. The bottom line is to make the ideas clear. 
    2. Sunday (Jan 18). After Hanshan finished replicating kdd'07 paper. We plan to finish TKDD paper for a submission. Haishan will complete the ontology documents (NEMOLex) ASAP, as well. 
    3. After Jan 18 to Febuary: Haishan will implement mapping discovery with some new data mining methods based on our initial ontologies. Dejing will help the process. We will use mapping results to refine and verify the ontologies.
/wiki/images/1/17/Fish1.png /wiki/images/e/ea/Fish2.png /wiki/images/f/fa/Fish3.png /wiki/images/f/ff/Fish4.png /wiki/images/4/40/Fish5.png /wiki/images/c/c5/Fish6.png
Personal tools