Table 4 The Average AUROC and Average AUPR with standard deviation in parentheses of different methods based on three types of networks, random (RN), small-world (SW) and scare-free (SF).

From: Reconstruction of Complex Directional Networks with Group Lasso Nonlinear Conditional Granger Causality

Type

N

〈k〉

σ

M

CGC

Lasso-CGC

NCGC

GLasso-NCGC

Aver-AUROC/Aver-AUPR (SD/SD)

RN

100

5

0.1

400

0.6367/0.2257

0.8118/0.6441

0.5472/0.1369

0.9315/0.7719

   

(0.0290/0.0395)

(0.0072/0.0181)

(0.0112/0.0075)

(0.0083/0.0227)

5

0.1

600

0.8022/0.4947

0.8989/0.7981

0.6321/0.2322

0.9755/0.9072

   

(0.0234/0.0405)

(0.0102/0.0213)

(0.0256/0.0317)

(0.0036/0.0128)

5

0.3

400

0.6518/0.2563

0.7963/0.5994

0.5530/0.1389

0.9186/0.7476

   

(0.0180/0.0114)

(0.0129/0.0320)

(0.0132/0.0042)

(0.0052/0.0102)

8

0.1

400

0.7394/0.3308

0.9281/0.8382

0.5966/0.1341

0.9714/0.8967

   

(0.0233/0.0319)

(0.0104/0.0162)

(0.0192/0.0188)

(0.0034/0.0150)

SW

100

5

0.1

400

0.8648/0.4658

0.9769/0.8823

0.6065/0.0929

0.9996/0.9895

   

(0.0213/0.0503)

(0.0044/0.0142)

(0.0185/0.0140)

(0.0001/0.0034)

5

0.1

600

0.8895/0.5384

0.9716/0.8826

0.6949/0.1684

0.9997/0.9919

   

(0.0311/0.0733)

(0.0070/0.0275)

(0.0279/0.0255)

(0.0001/0.0027)

5

0.3

400

0.8527/0.4206

0.9643/0.8342

0.6072/0.0894

0.9975/0.9570

   

(0.0145/0.0312)

(0.0073/0.0243)

(0.0113/0.0093)

(0.0004/0.0085)

8

0.1

400

0.6500/0.1970

0.9538/0.8791

0.5476/0.1030

0.9950/0.9312

   

(0.0325/0.0309)

(0.0082/0.0196)

(0.0180/0.0081)

(0.0012/0.0118)

SF

100

5

0.1

400

0.8304/0.4703

0.9358/0.8377

0.7680/0.2798

0.9492/0.8812

   

(0.0128/0.0278)

(0.0055/0.0082)

(0.0217/0.0320)

(0.0077/0.0174)

5

0.1

600

0.9216/0.6986

0.9498/0.8854

0.8939/0.5795

0.9638/0.9056

   

(0.0098/0.0183)

(0.0025/0.0047)

(0.0110/0.0515)

(0.0050/0.0112)

5

0.3

400

0.8464/0.4595

0.9233/0.8026

0.6929/0.1941

0.9443/0.8415

   

(0.0198/0.0270)

(0.0057/0.0099)

(0.0166/0.0150)

(0.0078/0.0100)

8

0.1

400

0.7144/0.3078

0.8322/0.6410

0.6409/0.1786

0.9095/0.7848

   

(0.0170/0.0206)

(0.0086/0.0211)

(0.0266/0.0279)

(0.0072/0.0230)

  1. The values are computed by averaging over 10 independent realizations. The results of different methods at different conditions are explored (type, N, 〈k〉, σ and M). Here, N is the size of network, 〈k〉 is average degree of network, σ is gaussian noise intensity, M is the number of samples. The highest scores of the Average AUROC and Average AUPR are highlighted.
Back to article page