Using callback for CrossVal

[ksehic]

Hi @QB3 @mathurinm

I was running your CrossVal example with callback function to get MSPE on test data and I got the error message

"ValueError: matmul: Input operand 1 does not have enough dimensions (has 0, gufunc core with signature (n?,k),(k,m?)->(n?,m?) requires 1)"

Thanks!

import numpy as np
import sklearn

from sklearn.linear_model import LassoCV
from sklearn.datasets import make_regression
from sklearn.model_selection import KFold
from sklearn.metrics import mean_squared_error
from sklearn.model_selection import train_test_split

from sparse_ho import ImplicitForward, grad_search
from sparse_ho.models import Lasso
from sparse_ho.criterion import HeldOutMSE, CrossVal
from sparse_ho.optimizers import LineSearch
from sparse_ho.utils import Monitor
from sparse_ho.utils_plot import discrete_cmap


dataset = 'simu'

if dataset == 'rcv1':
    X, y = fetch_libsvm('rcv1_train')
else:
    X, y = make_regression(
        n_samples=500, n_features=1000, noise=40,
        random_state=42)
        
X, X_test, y, y_test = train_test_split(
        X, y, test_size=0.15, random_state=0)

kf = KFold(n_splits=5, shuffle=True, random_state=42)

n_samples = len(y)
alpha_max = np.max(np.abs(X.T.dot(y))) / n_samples

n_alphas = 10
alphas = np.geomspace(alpha_max, alpha_max / 1_000, n_alphas)

tol = 1e-8
max_iter = 1e5

estimator = sklearn.linear_model.Lasso(
    fit_intercept=False, max_iter=1000, warm_start=True, tol=tol)

estimator = sklearn.linear_model.Lasso(
    fit_intercept=False, max_iter=1000, warm_start=True, tol=tol)
model = Lasso()
criterion = HeldOutMSE(None, None)
alpha0 = alpha_max / 10

objs_test = []


def callback(val, grad, mask, dense, alpha):
    # The custom quantity is added at each outer iteration:
    # here the prediction MSE on test data
    objs_test.append(mean_squared_error(X_test[:, mask] @ dense, y_test))

monitor_grad = Monitor(callback=callback)
cross_val_criterion = CrossVal(criterion, cv=kf)
algo = ImplicitForward()
optimizer = LineSearch(n_outer=10, tol=tol)
grad_search(
    algo, cross_val_criterion, model, optimizer, X, y, alpha0,
    monitor_grad)