Convolution

[EnLAI111]

Analysis:

• [ ] ADMM (quadratic) Problem: A.T A u + gamma D.T @ D @ u = A.T * y - gamma D.T @ z - D.T mu we want to obtain u from this expression.
• [ ] ChambollePock (quadratic, huber)
• [ ] CondaVu (quadratic) Problem: It can be also used for huber
• [ ] PGD (proxTV) (quadratic, huber) Problem: stepsize (avant we set it as 1.9 / np.linalg.norm(self.A, ord=2))
• [ ] Dual prox GD (quadratic) Problem: we can't get the inverse of A np.diff(np.linalg.pinv(A), axis=0) Problem: stepsize

Synthesis:

• [ ] ISTA (quadratic, huber) Problem: stepsize (avant we set it as 1.9 / np.linalg.norm(self.A @ L, ord=2))
• [ ] Celer (quadratic)
• [ ] skglm (quadratic, huber) Problem: self.clf.fit(AL, self.y)
• [ ] FP (quadratic) Problem: stepsize (avant we set it as 1.9 / (n * np.max((AL**2).sum(axis=1))))