Fast Likelihood Ratio Optimization Based Upon Genaralized Logarithm and Its Applications
| Project/Area Number |
22656088
|
|---|
| Research Category |
Grant-in-Aid for Challenging Exploratory Research
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Communication/Network engineering
|
|---|
| Research Institution | Waseda University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2010 – 2012
|
| Project Status |
Completed (Fiscal Year 2012)
|
| Budget Amount *help |
¥3,540,000 (Direct Cost: ¥3,000,000、Indirect Cost: ¥540,000)
Fiscal Year 2012: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2011: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2010: ¥1,200,000 (Direct Cost: ¥1,200,000)
|
|---|
| Keywords | 信号処理 / アルファＨＭＭ推定アルゴリズム / Ｒａｐｉｄ ＩＣＡアルゴリズム / 高速化 / アルファ対数尤度比 / 過去情報 / モーメンタム項 / 隠れマルコフモデル / 高速推定アルゴリズム / アルファHMMアルゴリズム / アルファEMアルゴリズム |
|---|
| Research Abstract |
Likelihood optimization for learning algorithms was generalized by using the alpha-logarithm. This generalization led to a faster convergence than traditional methods. Algorithms on hidden Markov model estimation and independent component analysis were chosen since they have high ramifications. The use of the alpha-logarithm appears as the utilization of past information via momentum terms. This property enabled faster convergence than traditional methods.
|
Report
(4 results)
Research Products
(24 results)
