I

Velocity

Velocity f

0.2 0.4 0.6 0.8 Observed values

0.2 0.4 0.6 0.8 Observed values

• Chironomus o Chironomidae □ Odonata ■ Baetis a Oligochaeta a Hirudinea o Gastropoda

Figure 9 Variation of output ranges in densities of selected taxa in benthic macroinvertebrate community when input values ranging +50% and -50% were provided to different environmental variables. (Variation was expressed as standard deviation of 11 observations.) Adapted from Chon T-S, Kwak I-S, Park Y-S, Kim T-H, and Kim YS (2001) Patterning and short-term predictions of benthic macroinvertebrate community dynamics by using a recurrent artificial neural network. Ecological Modelling 146:181-193, with permission from Elsevier.

network was feasible in patterning the sequential line movement of oligochaetes after the treatments of toxic substances. The body shape of the test specimens of L. variegatus was recorded by using an observation system consisting of an observation aquarium, a camera, and software for an image recognition system. During the observation period, groups and individuals of Lumbriculus were placed in a glass aquarium (diameter: 9 cm), and line positions were scanned from top view in 0.25 s intervals using a CCTV camera.

For input data, 13 x-y coordinates of the line shape of blackworm specimens were obtained through computer recognition system measured in every 0.25 s interval. The coordinates were converted to a line consisting of 12 subsegments with 12 lengths and 11 angles.

The line data of the specimens at the beginning point were selected for the initial data for each section. Subsequently, 11 more line data selected in every 25 s interval were merged to the initial line data. In total, the 12 sequential line data for 5 min (25 s x 12 = 300 s) were regarded as a sample unit provided to RSOM as input.

Clustering appeared in a characteristic manner according to Ward's linkage method, the samples were divided into six groups, with inclusion of sub- and sub-sub-clusters (Figures 10a-10c).

Although clustering was diverse, the gradient was still observed diagonally on the map. While the body segments with larger and less folded bodies were observed in cluster F in the area of bottom right, the shorter and strongly folded bodies were placed in cluster E in the area of top left (Figures 10a and 10b). Cluster F presented the specimens before the treatments, while cluster E was grouped by the specimens after the treatment. Clusters were further divided according to degree of contraction and folding. Cluster F, for instance, was divided to FI and FII. In cluster FI at the bottom-right corner, the longer and less curved specimens were observed. Subcluster FII was differentiated from FI with respect to the stronger degree of contraction and folding. The subclusters were further divided into sub-sub-clusters. While more straight forms were allowed in the body segments in sub-sub-cluster FIa, folding (e.g., the fifth snapshot in FIb, Figure 10b) was produced in the sequence of the line-movement in the sub-sub-cluster FIb.

E I:

S

- Ni

N> i ""

V

E II:

)

r )

r

*r*

F II a:

—S

<f

r —

F II b:

<r

c C

?

I N \

"N.

F I a:

—•

{ /

F I b:

i—■' !__■

r-

^ —> ^

/

t=

1

t=4

t

Figure 10 (a) Grouping of the sequential line movements of Lumbriculus specimens after training with RSOM based on angles and lengths of body segments (clustering carried out on the patterned nodes by the Ward linkage method). White and black circles indicate samples before and after copper treatments, respectively (max. number of the samples grouped in one unit: 100). (b) Time sequence of the line movements of Lumbriculus in different clusters. (c) Dendrogram of the RSOM units by Ward's linkage method. Adapted from Son K-H, Ji CW, Park Y-M, etal. (2006) Recurrent Self-Organizing Map implemented to detection of temporal line-movement patterns of Lumbriculus variegatus (Oligochaeta: Lumbriculidae) in response to the treatments of heavy metal. In: Kungolos AG, Brebbia CA, Samaras CP, Popov V (eds.) Environmental Toxicology. WIT Transaction on Biomedicine and Health, vol. 10, pp. 77-91. Southampton, UK: WIT Press.

Similarly, other subclusters were more finely divided. Overall, the training with RSOM demonstrated that the time-series line movement was accordingly clustered on the map of reduced dimension.

Was this article helpful?

0 0
10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook


Post a comment