Sato, M. and Ishii, S, On-line EM algorithm for the normalized Gaussian network, Neural Computation, Vol. 12, pp. 407--432, 2000. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Clustering Binary Data Streams with K-means - Ordonez (2003) (5 citations) (Correct)
....sensitivity to outliers and skewed distributions and converging to poor locally optimal solutions. This article introduces several improvements to K means to cluster binary data streams. The K means variants studied in this article include the standard version of K means [19, 9] Online K means [25], Scalable K means [6] and Incremental K means, a variant we propose. 1.1 Contributions and article outline This article presents several K means improvements to cluster binary data streams. Improvements include simple sufficient statistics for binary data, efficient distance computation for ....
M. Sato and S. Ishii. On-line EM algorithm for the normalized Gaussian network. Neural Computation, 12(2), 2000.
FREM: Fast and Robust EM Clustering for Large Data Sets - Ordonez, Omiecinski (2002) (Correct)
....are outliers making computations undefined or unstable. It may produce inaccurate results with high dimensional data. This is related to the problem of computing distances in high dimensional space [1, 4] There has been work on accelerating and improving EM, like incremental EM [22] On line EM [31, 34], EM for high dimensional basket data [27] and Scalable EM [6] but no solution solves all problems listed above. A survey of other related approaches from the Machine Learning community can be found in [20] 1.2 Contributions and article outline This article presents an improved clustering ....
....somewhere in the middle, but closer to the last scenario. That is, running the M step as few times as possible. This number of times will be related to k and d, but not to n. When a good approximation to the solution has been reached then we can run normal EM iterations to converge. On line EM [6, 31, 34] represents the fastest version that can be derived from EM, but without guarantees about optimality or stability of the solution. On line EM executes the M step after every E step for each point. So it only makes one scan over the data set. However, it is very sensitive to the order of points, ....
M. Sato and S. Ishii. On-line em algorithm for the normalized gaussian network. Neural Computation, 12(2), 2000.
Using navigation data to improve IR functions in the context.. - Hansen, Shriver (Correct)
....while the standard EM algorithm is sufficient for the relatively small data sets presented in this paper, it is known not to scale very well as either the number of queries or the number of query clusters increases. To combat this, we have started working with on line versions of the EM algorithm [13] that can process individual search sessions as they arrive. This requires a reformulation of the model as presented here, but we believe it will give reasonable performance. We are also extending our model to treat the query string as a collection of query terms and not simply a sorted list as we ....
M. Sato and S. Ishii. On-line EM algorithm for the normalized gaussian network. Neural Computation, 12(2):407--432, Feb. 2000.
Model-based Reinforcement Learning for Partially.. - With Sampling-Based State Self-citation (Ishii) (Correct)
No context found.
Sato, M. and Ishii, S, On-line EM algorithm for the normalized Gaussian network, Neural Computation, Vol. 12, pp. 407--432, 2000.
A Reinforcement Learning Scheme for a Multi-Agent Card.. - Ishii, Fujita.. (2005) (1 citation) Self-citation (Ishii) (Correct)
No context found.
M.Sato and S.Ishii. On-line EM algorithm for the normalized Gaussian network. Neural Computation, vol.12, no.2, pp.407-432, 2000.
A Reinforcement Learning Scheme for a Partially-Observable.. - Ishii, Fujita, Al. (2004) Self-citation (Ishii) (Correct)
No context found.
Sato, M., & Ishii, S. (2000). On-line em algorithm for the normalized gaussian network. Neural Computation, 12, 407--432.
Reconstruction of chaotic dynamics by on-line EM algorithm - Shin Ishiiy Masa-Aki Self-citation (Sato Ishii) (Correct)
....The NGnet is a network of local linear regression units. The model softly partitions the input space by using normalized Gaussian functions, and each local unit linearly approximates the output within its partition. The NGnet is trained by the on line EM algorithm, which we previously proposed [23]. It has been shown that this on line EM algorithm is faster than the gradient descent algorithm. Moreover, the on line EM algorithm is faster and more accurate than the batch EM algorithm [31] especially for a large amount of data [23, 25] Therefore, it is advantageous to use the online EM ....
....by the on line EM algorithm, which we previously proposed [23] It has been shown that this on line EM algorithm is faster than the gradient descent algorithm. Moreover, the on line EM algorithm is faster and more accurate than the batch EM algorithm [31] especially for a large amount of data [23, 25]. Therefore, it is advantageous to use the online EM algorithm for learning chaotic dynamics because a large amount data is necessary for approximating the vector field accurately. In the on line EM algorithm, the positions of the local units can be adjusted according to the input and output data ....
[Article contains additional citation context not shown here]
Sato, M., & Ishii, S. (2000). On-line EM algorithm for the normalized Gaussian network. Neural Computation, 12, 407-432.
Fast Learning of On-line EM Algorithm - Sato (58 citations) Self-citation (Sato) (Correct)
No context found.
Sato, M., & Ishii, S. (1999). On-line EM Algorithm for the Normalized Gaussian Network. To appear in Neural Computation.
Application of Reinforcement Learning to Balancing of Acrobot - Yoshimoto, Ishii, Sato (1999) (2 citations) Self-citation (Sato Ishii) (Correct)
....because the state and action spaces in these systems are continuous. In such cases, a good function approximator and a fast learning algorithm are crucial to achieving good performance. In our previous paper [2] we proposed an RL method based on our previously developed on line EM algorithm [3] and applied it to a couple of control problems in which the state and action spaces were continuous. In this article, we discuss applying this method to the task of how to control the balance of an acrobot [4, 5] that is a two link underactuated robot roughly analogous to a gymnast swinging on a ....
....the Qfunction that follows the current actor, and is based on the Bellman s equation [7] The actor and the critic are then approximated by the normalized Gaussian networks (NGnet s) 8] which are networks of local linear regression units. The NGnet s are trained by the on line EM algorithm [3]. The on line EM algorithm was chosen because it has been shown that it is faster than the gradient method and is suitable for dynamic environments in which the input output distribution of data changes over time. The approximation of the actor critic is one such problem. We also introduce ....
[Article contains additional citation context not shown here]
Sato, M., & Ishii, S., "On-line EM algorithm for the normalized gaussian network," Neural Computation, in press, 1999.
Critical Lines in Symmetry of Mixture Models and its.. - Fukumizu, Akaho, Amari (2002) (Correct)
No context found.
M. Sato and S. Ishii. On-line EM algorithm for the normalized Gaussian network. Neural Computation, 12(2):2209--2225, 2000.
Unsupervised Statistical Models for General Object Recognition - Carbonetto (2003) (Correct)
No context found.
Masa-aki Sato and Shin Ishii. On-line EM algorithm for the Normalized Gaussian Network. Neural Computation, 12(2):407--432, 2000.
Clustering Binary Data Streams with K-means - Carlos Ordonez Carlos (2003) (5 citations) (Correct)
No context found.
M. Sato and S. Ishii. On-line EM algorithm for the normalized Gaussian network. Neural Computation, 12(2), 2000.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC