Estimators setup
Reservoir TE distribution
Zdrojaky:
LorenzTEsig.py
LorenzTEnormalizedSig.py
Reservoir TE entropy continued
Lorenz prediction task parameter exploration
Initial network setup
Spectral radius
argmin NRMSE=0.2, min NRMSE=0.00001525
Win
argmin NRMSE=0.1, min NRMSE=0.00155
Orthogonalisation
argmin NRMSE=0, min NRMSE=0.001396
Orthonormalisation
argmin NRMSE=5, min NRMSE=0.01466
Všetko spolu
OG: argmin NRMSE=(0,0.2), min NRMSE=0.00001164
ON: argmin NRMSE=(10,0.125), min NRMSE=0.00001281
Prezentácia DP na Seminár z matematickej štatistiky
Prezentácia: pdf
Reservoir TE entropy
TE a Lorenz attractor
TE a predikcie X(t) zložky Lorenzov systému.
NRMSE
Mean TE
Mean TE reservoir-output
TE Input/Output
Zdrojaky:
LorenzTE.py
LorenzTEnormalized.py
LorenzTEnormalizedOG.py
LorenzTEnormalizedON.py
Reservoir embedding examination
Optimálne nastavenia parametrov embedding vektora podľa Ragwitz-Kantz kritéria pre reservoir. Tj. 
Initial network setup
History
Tau
Frequencies
Zdrojaky:
26: Transfer entropy, Active information storage & Lyapunov exponent in NARMA, Mackey-Glass prediction, Lorenz attractor & Memory capacity task
25: Transfer entropy, Active information storage & Lyapunov exponent in NARMA & Mackey-Glass prediction task
Pokúšal som sa zreprodukovať výsledky Boedeckera z Information Processing in Echo State Networks at the Edge of Chaos
Initial network setup
Estimators setup
Výsledky:
Transfer entropy vs. Lyapunov exponent
NRMSE vs. Lyapunov exponent
Active information storage vs. Lyapunov exponent
Zdrojaky:
NARMA prediction task parameter exploration
Initial network setup
Spectral radius
argmin NRMSE=0.95, min NRMSE=0.84877
Win
argmin NRMSE=0.8, min NRMSE=0.909755
Orthogonalisation
argmin NRMSE=66, min NRMSE=0.823972
Orthonormalisation
argmin NRMSE=69, min NRMSE=0.83421
Všetko spolu
OG: argmin NRMSE=(16,29), min NRMSE=0.8187
ON: argmin NRMSE=(29,29), min NRMSE=0.823497
+ Bias
argmin NRMSE=(19,28), min NRMSE=0.838228
+Bias + I/O
