Budget Amount *help |
¥13,910,000 (Direct Cost: ¥10,700,000、Indirect Cost: ¥3,210,000)
Fiscal Year 2013: ¥4,550,000 (Direct Cost: ¥3,500,000、Indirect Cost: ¥1,050,000)
Fiscal Year 2012: ¥4,550,000 (Direct Cost: ¥3,500,000、Indirect Cost: ¥1,050,000)
Fiscal Year 2011: ¥4,810,000 (Direct Cost: ¥3,700,000、Indirect Cost: ¥1,110,000)
|
Research Abstract |
We have improved a logic-based modeling language PRISM which unifies statistical machine learning and logical inference by adding a general MCMC (Markov chain Monte Carlo) method, VT (Viterbi training) and VB-VT that extends VT with variational Bayes. We also enabled PRISM to calculate an infinite sum of probabilities through solving probability equations, which is applied to intention recognition of users from web log session data.
|