1887
banner image
No data available.
Please log in to see this content.
You have no subscription access to this content.
No metrics data to plot.
The attempt to load metrics for this article has failed.
The attempt to plot a graph for these metrics has failed.
The architecture of dynamic reservoir in the echo state network
Rent:
Rent this article for
USD
10.1063/1.4746765
/content/aip/journal/chaos/22/3/10.1063/1.4746765
http://aip.metastore.ingenta.com/content/aip/journal/chaos/22/3/10.1063/1.4746765
View: Figures

Figures

Image of FIG. 1.
FIG. 1.

System architecture of ESN.

Image of FIG. 2.
FIG. 2.

System architectures of the new ESNs

Image of FIG. 3.
FIG. 3.

DR architecture of MESN.

Image of FIG. 4.
FIG. 4.

The training and testing NRMSE of conventional ESN, NWESN, BAESN, and MESN on dataset 1 with and dataset 2. The y-axis represents NRMSE, and the x-axis represents the reservoir size. The reservoir size varies from 50 to 500 with step size 50; (a) the training NRMSE, (b) the testing NRMSE.

Image of FIG. 5.
FIG. 5.

The testing NRMSE versus spectral radius on datasets 1 and 2, respectively. The y-axis represents NRMSE, and the x-axis represents the spectrum radius. The range of spectral radius is set to [0.1, 2.5] with step size 0.1.

Image of FIG. 6.
FIG. 6.

The testing NRMSE versus spectral radius on dataset 1; (a) testing on classical ESN with spectral radius range [0, 1.5]; (b) testing on NWESN with spectral radius range [0, 2]; (c) testing on BAESN with spectral radius range [0, 2]; and (d) testing on MESN with spectral radius range [0, 5].

Image of FIG. 7.
FIG. 7.

The degree distributions of reservoirs; (a) DR in traditional ESN; (b) DR in NWESN; (c) DR in BAESN; and (d) DR in MESN.

Image of FIG. 8.
FIG. 8.

The forgetting curves; (a) the forgetting curve of the conventional ESN with a reservoir connectivity = 0.2; (b) the forgetting curve of NWESN; (c) the forgetting curve of BAESN; (d) the forgetting curve of MESN model with 1 scale-free group and 2 small-world groups.

Image of FIG. 9.
FIG. 9.

Testing NRMSE of different ratio between scale-free groups and small-world groups on dataset 1. The x-axis represents the ratio between the scale-free groups and the small-world groups, and the y-axis represents NRMSE. Parts (a)–(e) denote the reservoir with 5, 10, 15, 20, and 25 groups, respectively.

Image of FIG. 10.
FIG. 10.

Testing NRMSE of different ratio between scale-free groups and small-world groups on dataset 2. The x-axis represents the ratio between scale-free groups and small-world groups. The y-axis represents NRMSE. Parts (a)–(e) denote the reservoir with 5, 10, 15, 20, and 25 groups, respectively.

Image of FIG. 11.
FIG. 11.

The short-term memory capacity of MESN with different reservoir size and different ratios. The x-axis represents ratios between scale-free groups and small-world groups, and the y-axis represents the MC value; (a) DR size = 100, group number = 5; (b) DR size = 200, group number = 10; (c) DR size = 200, group number = 5; and (d) DR size = 400, group number = 10.

Image of FIG. 12.
FIG. 12.

The influence of parameters s and p in NWESN. The x-axis represents s, the y-axis represents p, and the z-axis represents the testing NRMSE. Therange of p is set to [0.1, 0.9] with step size 0.1. The range of s is set to [2,20] with step size 0.1. The dynamic reservoir size is 100, 200, and 300, respectively.

Image of FIG. 13.
FIG. 13.

The influence of parameter m on BAESN. The x-axis represents m and the y-axis represents NRMSE. The range of m is set to [0, 20] with step size 2. The dynamic reservoir size is 100, 200, and 300, respectively.

Loading

Article metrics loading...

/content/aip/journal/chaos/22/3/10.1063/1.4746765
2012-08-20
2014-04-16
Loading

Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: The architecture of dynamic reservoir in the echo state network
http://aip.metastore.ingenta.com/content/aip/journal/chaos/22/3/10.1063/1.4746765
10.1063/1.4746765
SEARCH_EXPAND_ITEM