Table 1 Performance on tasks and capacity measures for different dynamical systems. Given as Cohen’s kappa score for XOR, tXOR and XORXOR, maximum delay up to which accuracy is above chance level for delayed classification, and squared correlation coefficient for NARMA5.

From: A refined information processing capacity metric allows an in-depth analysis of memory and nonlinearity trade-offs in neurocomputational systems

 

ESN

\(\rho\) 0.9

FPUT

\(\alpha\) 0.25

BRN

distr. p 1

BRN

spatial \(\sigma\) 20

MC

distr. p 1

MC

spatial \(\sigma\) 20

Readout units

50

64

1000

1000

447

447

XOR

1.

0.85

1.

0.99

1.

0.99

XORXOR

0.73

 

0.17

0.62

0.21

0.22

tXOR

 

0.72

1.

0.99

1.

0.99

Classification

  

26

18

42

50

NARMA5

 

0.88

0.29

0.2

0.43

0.17

Capacity

49

55

38

100

49

40

Max. degree

11

3

35

285

71

255

Max. delay

69

688

32

10

53

14