How to predict on unknown data about reservoirpy HOT 10 CLOSED

Hello @wilbertmatthew

Thank you for using reservoirpy.
Please make your question more explicit, you are asking several questions at the same time, and put link to the line code where you have an issue if you have one.
Thanks

from reservoirpy.

Thanks for your response.

Hope this clarifies my issue:

Issue 1: I train on 1-2000 values of the dataset and predict on 4000-4100 values of the dataset using code y_pred = model.fit(X, y).run(X_test). That works.

How would I run the app again, with same dataset or another dataset with a similar signal pattern, but only perform a prediction without training? That is can I save the model and prove that prediction is working on different datasets than the dataset used for training.

import matplotlib.pyplot as plt
import numpy as np
from reservoirpy.nodes import NVAR, Ridge
from reservoirpy.datasets import lorenz
from reservoirpy.datasets import mackey_glass
from reservoirpy.nodes import Input
import reservoirpy as rpy
rpy.verbosity(1)

nvar = NVAR(delay=2, order=2, strides=1)
readout = Ridge(1, ridge=2.5e-6)
model = nvar >> readout

tau = 17
data = mackey_glass(10000, tau=tau)
data = data.reshape(-1, 1)

VERBOSE = True

train_size = 2000
test_size = 2000
horizon = 1 # horizon p of the forecast (predict X[t+p] from X[t])

X = data[:train_size]
y = data[horizon : train_size + horizon]
X_test = data[train_size : train_size + test_size]
y_test = data[train_size + horizon : train_size + test_size + horizon]

normalize = True

if VERBOSE:
print("X, y dimensions", X.shape, y.shape)
print("X_test, y_test dimensions", X_test.shape, y_test.shape)

y_pred = model.fit(X, y).run(X_test)

plt.figure(figsize=(12, 4))
plt.plot(y_pred, color="red", lw=1.5, label="Predictions")
plt.plot(y_test, color="blue", lw=0.75, label="Ground truth")
plt.title("Output predictions against real timeseries")
plt.legend()
plt.show()

from reservoirpy.

How would I run the app again, with same dataset or another dataset with a similar signal pattern, but only perform a prediction without training?

since you already trained the model:

y_pred = model.fit(X, y).run(X_test)

you just need to do:
y_pred_newdata = model.run(X_test_newdata)

Does it answer your question?

from reservoirpy.

from reservoirpy.

from reservoirpy.

Bonjour,
Comment importer la série de Lorenz 3d pour faire une prédiction

from reservoirpy.

Hello @Mervira,

You can just import it from the datasets module, please refer to documentation for more information.

from reservoirpy.

from reservoirpy.

from reservoirpy.

Hello @Mervira,
Please open new issues when you have unrelated questions. I will try answer both here for now:

1. Your first question is a matplotlib (or whatever plotting tool you use) problem. If you want 3 different plots, simply create 3 different figures, one with each variable of the dataset (x[:, 0], x[:, 1] and x[:, 2], x being your Lorenz series array)
2. Your second question is raspberry problem. If your raspberry has an operating system with Python 3.8 or higher installed on it, then it should work. Raspberry are just tiny computers.

from reservoirpy.