Abstract
We develop new techniques for time series classification based on hierarchical Bayesian generative models (called mixed-effect models) and the Fisher kernel derived from them. A key advantage of the new formulation is that one can compute the Fisher information matrix despite varying sequence lengths and varying sampling intervals. This avoids the commonly-used ad hoc replacement of the Fisher information matrix with the identity which destroys the geometric invariance of the kernel. Our construction retains the geometric invariance, resulting in a kernel that is properly invariant under change of coordinates in the model parameter space. Experiments on detecting cognitive decline show that classifiers based on the proposed kernel out-perform those based on generative models and other feature extraction routines, and on Fisher kernels that use the identity in place of the Fisher information.
| Original language | English (US) |
|---|---|
| Title of host publication | Advances in Neural Information Processing Systems 21 - Proceedings of the 2008 Conference |
| Pages | 1961-1968 |
| Number of pages | 8 |
| State | Published - 2009 |
| Event | 22nd Annual Conference on Neural Information Processing Systems, NIPS 2008 - Vancouver, BC, Canada Duration: Dec 8 2008 → Dec 11 2008 |
Other
| Other | 22nd Annual Conference on Neural Information Processing Systems, NIPS 2008 |
|---|---|
| Country | Canada |
| City | Vancouver, BC |
| Period | 12/8/08 → 12/11/08 |
Fingerprint
ASJC Scopus subject areas
- Information Systems
Cite this
Hierarchical Fisher kernels for longitudinal data. / Lu, Zhengdong; Leen, Todd K.; Kaye, Jeffrey.
Advances in Neural Information Processing Systems 21 - Proceedings of the 2008 Conference. 2009. p. 1961-1968.Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Hierarchical Fisher kernels for longitudinal data
AU - Lu, Zhengdong
AU - Leen, Todd K.
AU - Kaye, Jeffrey
PY - 2009
Y1 - 2009
N2 - We develop new techniques for time series classification based on hierarchical Bayesian generative models (called mixed-effect models) and the Fisher kernel derived from them. A key advantage of the new formulation is that one can compute the Fisher information matrix despite varying sequence lengths and varying sampling intervals. This avoids the commonly-used ad hoc replacement of the Fisher information matrix with the identity which destroys the geometric invariance of the kernel. Our construction retains the geometric invariance, resulting in a kernel that is properly invariant under change of coordinates in the model parameter space. Experiments on detecting cognitive decline show that classifiers based on the proposed kernel out-perform those based on generative models and other feature extraction routines, and on Fisher kernels that use the identity in place of the Fisher information.
AB - We develop new techniques for time series classification based on hierarchical Bayesian generative models (called mixed-effect models) and the Fisher kernel derived from them. A key advantage of the new formulation is that one can compute the Fisher information matrix despite varying sequence lengths and varying sampling intervals. This avoids the commonly-used ad hoc replacement of the Fisher information matrix with the identity which destroys the geometric invariance of the kernel. Our construction retains the geometric invariance, resulting in a kernel that is properly invariant under change of coordinates in the model parameter space. Experiments on detecting cognitive decline show that classifiers based on the proposed kernel out-perform those based on generative models and other feature extraction routines, and on Fisher kernels that use the identity in place of the Fisher information.
UR - http://www.scopus.com/inward/record.url?scp=84858767817&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84858767817&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:84858767817
SN - 9781605609492
SP - 1961
EP - 1968
BT - Advances in Neural Information Processing Systems 21 - Proceedings of the 2008 Conference
ER -