Describe the feature or idea you want to propose

This enhancement proposes the implementation of a new RecurrentRegressor that leverages a recurrent neural network (RNN) architecture for deep learning-based time series regression. The RNN architecture will be based on the existing RNNNetwork module defined in the networks subpackage. This model will be designed to learn temporal dependencies and sequential patterns in multivariate time series data, improving predictive performance on tasks where past information is critical.

The new estimator should be compatible with the deep learning API and should integrate seamlessly with the existing aeon framework.

Describe your proposed solution

The RecurrentRegressor will be implemented as a new class that inherits from BaseDeepRegressor. The following methods will be implemented:

• build_network(input_shape):
  Constructs the RNN model by instantiating the RNNNetwork class, configuring the number of layers, hidden dimensions, and activation functions according to the specified hyperparameters.

• fit(X, y):
  Trains the RNN model on the input time series X and targets y. It should include data preprocessing, loss computation, and backpropagation steps.

• get_test_params():
  Returns a dictionary of test parameters for unit testing. This ensures model correctness and integration within the testing pipeline.

Describe alternatives you've considered, if relevant

No response

Additional context

No response