https://www.kaggle.com/datasets/fedesoriano/heart-failure-prediction

From 11 features(Age, Sex, Cholesterol...), predict the chance of heart failure in pacients.

Since there are many features, the mathametical function that describes it a non linear differencial equation

Neural networks can act as universial approximators for complex non linear differencial equations

Mathematically, this is called the Universal approximation theorem

In the notebook, essecially, we are building an universally approximator based on those 11 features that predict the chance of heart failure.

PyTorch lightning was the main library used, because of its simplicicy in scaling the size of the model

Addicionally, it allows very easily the creation of the training loop and evaluation loop

Making the full creation of the model and the functions that train and eavluate it in asingle python class, in a single cell

The last reason was that, as of

Converting the dataset in tensors in the right format for training in the Bi direcional LSTM model

Finding the best Hyperparameters for the model

pip install pandas numpy ipykernel notebook scikit-learn torch lightning ipynbname

If you haven't already created a virtual environment for your project, you can do so using virtualenv or venv. Here's an example using venv:

python -m venv myenv

Replace myenv with the desired name for your virtual environment.

On Windows, activate the virtual environment using:

myenv\Scripts\activate

On macOS and Linux, use:

source myenv/bin/activate

Replace myenv with the name of your virtual environment.

Once the virtual environment is activated, you can install Jupyter Notebook using pip:

pip install jupyter

This will install Jupyter Notebook within your virtual environment.

To verify that Jupyter Notebook is installed in your virtual environment, you can run:

jupyter --version

This should display the version of Jupyter Notebook installed within your virtual environment.

You need to create a Jupyter Notebook kernel that is associated with your virtual environment. This allows you to use the packages installed in your virtual environment within Jupyter Notebook.

pip install ipykernel

python -m ipykernel install --user --name=myenv --display-name="name"

Replace myenv with the name of your virtual environment and choose a suitable display name.

Now, you can start Jupyter Notebook from within your virtual environment:

jupyter notebook

This will open a new Jupyter Notebook session in your web browser, and you should be able to select the "My Virtual Environment" kernel when creating a new notebook. This kernel will use the packages installed in your virtual environment.

It can be used as foundation for other time series classification models.

With quantization, the model gets a lot smaller without pratically losing accuracy

By being smaller, itcan be applied in sensores and small devices like IoT or microcontrolers

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Vivamus feugiat fringilla eros nec mattis. Cras nec sagittis risus, vel mattis odio. Sed erat massa, commodo nec rutrum ac, tincidunt quis magna. Pellentesque non tristique ante. Phasellus convallis ante tincidunt lacus tempor aliquam. Donec quis ipsum laoreet, pretium ligula quis, pulvinar ante. Nam fringilla nunc in accumsan tempus.