write a code Working on the same data, use linear regression to predict the solubility (target variable in the 'Solubility' column) using the previously calculated descriptors as features, by finding optimal weights and biases. Write and a batch gradient descent code for this, either using jax or numpy. Compare learning rates of 1e-3, 1e-6, 1e-9 when using the sum of squared errors or the mean squared error as loss function. Elaborate on what you observe by plotting (or trying to plot) the different learning curves