KoopmanESN/TrajTrans/StateOnly/ESN/KoopmanESNMain.py

import torch
from KoopmanESNModel import KoopmanESNConfig, KoopmanESN
from BaseModel.ESNModel import ESN
from DataProcess import DataLoad
import numpy as np
import matplotlib.pyplot as plt


KoopmanModel = torch.load('./ModelLib/DeepKoopmanModel.pt', map_location="cpu")
encoder = KoopmanModel.encoder
decoder = KoopmanModel.decoder
A = KoopmanModel.A.detach().numpy()
state_dim = KoopmanModel.state_dim
latent_dim = KoopmanModel.latent_dim

config = KoopmanESNConfig(
    units=latent_dim,
    lr=0.5,
    sr=0.9,
    sp=0.1,
    ridge=1e-7,
    train_start=19500,
    train_len=500,
    train_warmup=10,
    predict_warmup=100,
    predict_len=20,
    state_dim=latent_dim
)

# config = KoopmanESNConfig(
#     units=latent_dim,
#     lr=0.5,
#     sr=0.9,
#     sp=0.1,
#     ridge=1e-7,
#     train_start=10000,
#     train_len=500,
#     train_warmup=10,
#     predict_warmup=100,
#     predict_len=20,
#     state_dim=latent_dim
# )

KoopESNModel = KoopmanESN(
    config=config,
    A=A
)
ESNModel = ESN(
    config=config
)

_, DataTrain, DataWarm, DataVal = DataLoad('./DataLib/LatentNor.csv', config)
# _, DataTrain, DataWarm = DataLoad('./DataLib/DataNor.csv', config)

# KoopmanESN
KoopESNModel.koopmanesn_train(DataTrain)
DataPreKoop = KoopESNModel.predict(DataWarm)

# ESN
ESNModel.esn_train(DataTrain)
DataPreESN = ESNModel.predict(DataWarm)

DataPreKoop_Ten = torch.from_numpy(DataPreKoop).float()
DataPreESN_Ten = torch.from_numpy(DataPreESN).float()
DataVal_Ten = torch.from_numpy(DataVal).float()
DataWarm_Ten = torch.from_numpy(DataWarm).float()

StatePreKoop_Ten = decoder(DataPreKoop_Ten)
StatePreESN_Ten = decoder(DataPreESN_Ten)
StateVal_Ten = decoder(DataVal_Ten)
StateWarm_Ten = decoder(DataWarm_Ten)

StatePreKoop = StatePreKoop_Ten.detach().numpy()
StatePreESN = StatePreESN_Ten.detach().numpy()
StateVal = StateVal_Ten.detach().numpy()
StateWarm = StateWarm_Ten.detach().numpy()

KoopMSE = np.linalg.norm(StatePreKoop-StateVal, ord='fro')**2/np.prod(StateVal.shape)
ESNMSE = np.linalg.norm(StatePreESN-StateVal, ord='fro')**2/np.prod(StateVal.shape)
print(KoopMSE)
print(ESNMSE)

t = np.arange(config.predict_warmup + config.predict_len)

for fea in np.arange(state_dim):
    plt.figure(fea)
    plt.plot(t[:config.predict_warmup], StateWarm[:, fea],
             linestyle="-", color='black', label='StateWarm')
    plt.plot(t[-config.predict_len:], StatePreKoop[:, fea],
             linestyle="--", color='red', label='KoopmanESNPre')
    plt.plot(t[-config.predict_len:], StatePreESN[:, fea],
             linestyle="--", color='green', label='ESNPre')
    plt.plot(t[-config.predict_len:], StateVal[:, fea],
             linestyle="-", color='blue', label='StateReal')
    plt.legend()
    plt.show()