Jonny
@jchirinosrguez.bsky.social
I got a PhD in applied math not so long ago. I like inverse problems and optimization. I also drink a lot of coffe ☕
Indeed, a natural question arises: is such an interval nonempty? Actually, we show that this is the case whenever the noise level is small compared to the amplitude of the nonzero components of the true solution :)
Enjoy and thanks for reading!
4/n, n=4
Enjoy and thanks for reading!
4/n, n=4
November 20, 2025 at 8:56 PM
Indeed, a natural question arises: is such an interval nonempty? Actually, we show that this is the case whenever the noise level is small compared to the amplitude of the nonzero components of the true solution :)
Enjoy and thanks for reading!
4/n, n=4
Enjoy and thanks for reading!
4/n, n=4
Specifically, in both cases our result simply writes: there exists an interval of regularisation parameters,
λ \in (Θ,Ω),
under which the oracle solution is the unique global minimizer of the exact relaxation (and so of the original L0 problem), and is also the sparsest critical point.
3/n
λ \in (Θ,Ω),
under which the oracle solution is the unique global minimizer of the exact relaxation (and so of the original L0 problem), and is also the sparsest critical point.
3/n
November 20, 2025 at 8:54 PM
Specifically, in both cases our result simply writes: there exists an interval of regularisation parameters,
λ \in (Θ,Ω),
under which the oracle solution is the unique global minimizer of the exact relaxation (and so of the original L0 problem), and is also the sparsest critical point.
3/n
λ \in (Θ,Ω),
under which the oracle solution is the unique global minimizer of the exact relaxation (and so of the original L0 problem), and is also the sparsest critical point.
3/n
In particular, we study exact support recovery results for both least-squares and the KL divergence!
Our work is based on a novel quantity we propose, termed Bergman Restricted Strong Convexity, that generalizes well-known quantities out there such as the Restricted Strong Convexity, and LRIP!
2/n
Our work is based on a novel quantity we propose, termed Bergman Restricted Strong Convexity, that generalizes well-known quantities out there such as the Restricted Strong Convexity, and LRIP!
2/n
November 20, 2025 at 8:49 PM
In particular, we study exact support recovery results for both least-squares and the KL divergence!
Our work is based on a novel quantity we propose, termed Bergman Restricted Strong Convexity, that generalizes well-known quantities out there such as the Restricted Strong Convexity, and LRIP!
2/n
Our work is based on a novel quantity we propose, termed Bergman Restricted Strong Convexity, that generalizes well-known quantities out there such as the Restricted Strong Convexity, and LRIP!
2/n
Amazing initiative! Congrats!
November 29, 2024 at 10:17 PM
Amazing initiative! Congrats!