Issues with ESNClassifier and ESNRegresssor

[pavankumar-dss]

Hi, I'm exploring this tool as a part of my research on RCN and wanted to see if this would help me learn and test for my application. But so far I'm unable to get anywhere with this. I'm currently using the dev branch as the main branch is also similar issues. I have all the required pip packages installed and yet I'm unable to run the provided examples. Most of the time the issue lie within the ESNClassifier and ESNRegressor classes. I'm attaching the error logs below. I'm sorry, that I couldn't contribute more, it's because I'm not very familiar with Python and Machine Learning. Hopefully these logs should help to fix the issues.

From setup_local.ipynb

from sklearn.datasets import load_iris, load_digits
from sklearn.preprocessing import LabelBinarizer
from sklearn.model_selection import train_test_split
from pyrcn.extreme_learning_machine import ELMRegressor


def test_iris():
    X, y = load_iris(return_X_y=True)
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.05)
    lb = LabelBinarizer().fit(y)
    y_train_numeric = lb.transform(y_train)
    classifier = ELMClassifier(hidden_layer_size=10)
    classifier.fit(X_train, y_train_numeric)
    y_predicted_numeric = classifier.predict(X_test)
    y_predicted = lb.inverse_transform(y_predicted_numeric)

    for record in range(len(y_test)):
        print('predicted: {0} \ttrue: {1}'.format(y_predicted[record], y_test[record]))
        

test_iris()

Error Log:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[4], line 22
     18     for record in range(len(y_test)):
     19         print('predicted: {0} \ttrue: {1}'.format(y_predicted[record], y_test[record]))
---> 22 test_iris()

Cell In[4], line 14, in test_iris()
     12 y_train_numeric = lb.transform(y_train)
     13 classifier = ELMClassifier(hidden_layer_size=10)
---> 14 classifier.fit(X_train, y_train_numeric)
     15 y_predicted_numeric = classifier.predict(X_test)
     16 y_predicted = lb.inverse_transform(y_predicted_numeric)

File ~\Documents\Code\ReservoirComputing\PyRCN\src\pyrcn\extreme_learning_machine\_elm.py:500, in ELMClassifier.fit(self, X, y, n_jobs, transformer_weights)
    478 def fit(self, X: np.ndarray, y: np.ndarray,
    479         n_jobs: Union[int, np.integer, None] = None,
    480         transformer_weights: Optional[np.ndarray] = None) -> ELMClassifier:
    481     """
    482     Fit the classifier.
    483 
   (...)    498     self : Returns a trained ELMClassifier model.
    499     """
--> 500     self._validate_data(X, y, multi_output=True)
    501     self._encoder = LabelBinarizer().fit(y)
    502     super().fit(X, self._encoder.transform(y), n_jobs=n_jobs,
    503                 transformer_weights=None)

AttributeError: 'ELMClassifier' object has no attribute '_validate_data'

Another example:

From stock-price-prediction.ipynb

Echo State Network preparation

base_input_to_nodes = InputToNode(hidden_layer_size=100, input_activation='identity',
                                  k_in=1, input_scaling=0.6, bias_scaling=0.0)
base_nodes_to_nodes = NodeToNode(hidden_layer_size=100, spectral_radius=0.9, leakage=1.0, k_rec=10)

esn = ESNRegressor(input_to_node=base_input_to_nodes,
                   node_to_node=base_nodes_to_nodes,
                   regressor=IncrementalRegression(alpha=1e-8), random_state=10)

Training and Prediction.

X_train = X[0:train_len, :]
y_train = X[0+1:train_len+1, :]
X_test = X[0:train_len+future_total - future_len, :]
y_test = X[future_len:train_len+future_total, :]

esn.fit(X=X_train, y=y_train.ravel())
y_train_pred = esn.predict(X=X_train)
y_test_pred = esn.predict(X=X_test)

Error log:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[9], line 6
      3 X_test = X[0:train_len+future_total - future_len, :]
      4 y_test = X[future_len:train_len+future_total, :]
----> 6 esn.fit(X=X_train, y=y_train.ravel())
      7 y_train_pred = esn.predict(X=X_train)
      8 y_test_pred = esn.predict(X=X_test)

File ~\Documents\Code\ReservoirComputing\PyRCN\src\pyrcn\echo_state_network\_esn.py:308, in ESNRegressor.fit(self, X, y, n_jobs, transformer_weights)
    285 def fit(self, X: np.ndarray, y: np.ndarray,
    286         n_jobs: Union[int, np.integer, None] = None,
    287         transformer_weights: Optional[np.ndarray] = None) -> ESNRegressor:
    288     """
    289     Fit the regressor.
    290 
   (...)    306     self : Returns a trained ESNRegressor model.
    307     """
--> 308     self._validate_hyperparameters()
    309     if self.requires_sequence == "auto":
    310         self._check_if_sequence(X, y)

File ~\Documents\Code\ReservoirComputing\PyRCN\src\pyrcn\echo_state_network\_esn.py:427, in ESNRegressor._validate_hyperparameters(self)
    423     raise ValueError('Invalid value for requires_sequence, got {0}'
    424                      .format(self._requires_sequence))
    426 if not is_regressor(self._regressor):
--> 427     raise TypeError("The last step should be a regressor and "
    428                     "implement fit and predict '{0}' (type {1})"
    429                     "doesn't".format(self._regressor,
    430                                      type(self._regressor)))

TypeError: The last step should be a regressor and implement fit and predict 'IncrementalRegression(alpha=1e-08)' (type <class 'pyrcn.linear_model._incremental_regression.IncrementalRegression'>)doesn't

Even the intro notebook is having issues with the InputToNode, NodeToNode and NodeToOutput classes