Mixing
Is there a regularized variant of the Sherman-Morrison formula?
Clash Royale CLAN TAG#URR8PPP
up vote
0
down vote
favorite
Wikipedia has a good article on it so I won't bother rehashing it. I am trying to use it to do iterative updates of a precision matrix for a NN and it cannot go far without blowing up. This is a pity as the natural gradient method I am trying to adapt it for would be a lot more efficient if it worked.
For such tasks it should be better to do an update like $(A + u v^T + epsilon I)^-1$ where $epsilon$ is a very small constant. If it were possible, it would solve a lot of my problems. But since at the same time it is also so obvious and the fact that I cannot find anything on it makes it highly likely that others have tried and failed at making such an update.
Has there been any successful research on this?
linear-algebra
asked Jul 23 at 11:18
Marko Grdinic
1369
add a comment |Â
up vote
0
down vote
favorite
Wikipedia has a good article on it so I won't bother rehashing it. I am trying to use it to do iterative updates of a precision matrix for a NN and it cannot go far without blowing up. This is a pity as the natural gradient method I am trying to adapt it for would be a lot more efficient if it worked.
For such tasks it should be better to do an update like $(A + u v^T + epsilon I)^-1$ where $epsilon$ is a very small constant. If it were possible, it would solve a lot of my problems. But since at the same time it is also so obvious and the fact that I cannot find anything on it makes it highly likely that others have tried and failed at making such an update.
Has there been any successful research on this?
linear-algebra
asked Jul 23 at 11:18
Marko Grdinic
1369
add a comment |Â
up vote
0
down vote
favorite
up vote
0
down vote
favorite
Wikipedia has a good article on it so I won't bother rehashing it. I am trying to use it to do iterative updates of a precision matrix for a NN and it cannot go far without blowing up. This is a pity as the natural gradient method I am trying to adapt it for would be a lot more efficient if it worked.
For such tasks it should be better to do an update like $(A + u v^T + epsilon I)^-1$ where $epsilon$ is a very small constant. If it were possible, it would solve a lot of my problems. But since at the same time it is also so obvious and the fact that I cannot find anything on it makes it highly likely that others have tried and failed at making such an update.
Has there been any successful research on this?
linear-algebra
asked Jul 23 at 11:18
Marko Grdinic
1369
Wikipedia has a good article on it so I won't bother rehashing it. I am trying to use it to do iterative updates of a precision matrix for a NN and it cannot go far without blowing up. This is a pity as the natural gradient method I am trying to adapt it for would be a lot more efficient if it worked.
For such tasks it should be better to do an update like $(A + u v^T + epsilon I)^-1$ where $epsilon$ is a very small constant. If it were possible, it would solve a lot of my problems. But since at the same time it is also so obvious and the fact that I cannot find anything on it makes it highly likely that others have tried and failed at making such an update.
Has there been any successful research on this?
linear-algebra
asked Jul 23 at 11:18
Marko Grdinic
1369
asked Jul 23 at 11:18
Marko Grdinic
1369
asked Jul 23 at 11:18
Marko Grdinic
1369
asked Jul 23 at 11:18
Marko Grdinic
1369
1369
add a comment |Â
add a comment |Â
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f2860239%2fis-there-a-regularized-variant-of-the-sherman-morrison-formula%23new-answer', 'question_page');
);
Post as a guest
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password