implement hosvd

[dmdunla]

Now that ttensor is available, we can implement HOSVD

[ntjohnson1]

Sketched out the initial core functionality https://github.com/ntjohnson1/pyttb/tree/hosvd should be able to finish it out in the next week or two. Mostly just need to do some more numerical testing to sanity check the mapping changes between matlab and python. If someone needs it sooner they are welcome to build off the base.

[dmdunla]

I was planning to work on this next, but I need to finish catching up. Let me take a crack at this and then I’ll request a review. I’d like to create the tests that match and/or are derived from MATLAB results at the start of each new addition so we can validate implementations with a known output. I’ll add the tests first. I noticed in tucker_als the was use of numbers.Real. Is this the direction we should be going for all implementations of numbers in pyttb?

…

________________________________ From: ntjohnson1 ***@***.***> Sent: Saturday, March 4, 2023 12:11:42 PM To: sandialabs/pyttb ***@***.***> Cc: Dunlavy, Daniel M ***@***.***>; Author ***@***.***> Subject: [EXTERNAL] Re: [sandialabs/pyttb] implement hosvd (Issue #56) Sketched out the initial core functionality https://github.com/ntjohnson1/pyttb/tree/hosvd<https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fntjohnson1%2Fpyttb%2Ftree%2Fhosvd&data=05%7C01%7Cdmdunla%40sandia.gov%7Cc9c113bc3fb348978c9d08db1ce45872%7C7ccb5a20a303498cb0c129007381b574%7C1%7C0%7C638135539292998673%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=faSzSxzZnhgpPu0Htp7hzihKlAFLgjBYzcpDHWZfRVw%3D&reserved=0> should be able to finish it out in the next week or two. Mostly just need to do some more numerical testing to sanity check the mapping changes between matlab and python. If someone needs it sooner they are welcome to build off the base. — Reply to this email directly, view it on GitHub<https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fsandialabs%2Fpyttb%2Fissues%2F56%23issuecomment-1454845612&data=05%7C01%7Cdmdunla%40sandia.gov%7Cc9c113bc3fb348978c9d08db1ce45872%7C7ccb5a20a303498cb0c129007381b574%7C1%7C0%7C638135539292998673%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=5HBfzwr8Kbxg6Ik4wZ6ZSaxtC%2B02FphJeH9y%2B29GDdU%3D&reserved=0>, or unsubscribe<https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAHIY2IQQRJLINAS6OFD7GHTW2OHW5ANCNFSM6AAAAAAVEOHJDY&data=05%7C01%7Cdmdunla%40sandia.gov%7Cc9c113bc3fb348978c9d08db1ce45872%7C7ccb5a20a303498cb0c129007381b574%7C1%7C0%7C638135539292998673%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=SYJ8AZoB%2F%2FLRkr1dgbGKdGl%2BGNp8GPq0mw%2FQsxjiNOA%3D&reserved=0>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[nick-cerebras]

Sounds good. I'll hold off then.

I was exploring going in the directions of numbers.Real since it seemed to be a more general solution for detecting generic scalars, but unfortunately it doesn't seem to be compliant with mypy type validation. Currently across pyttb we test for numeric scalars in a variety of different ways. Eventually we probably just centralize some of that to a utility to make sure we stay consistent. I think we can just use float but am not sure if there was some edge case why we were checking for so many other things before.

[dmdunla]

From what you saw, and because I do t want to dig right now (but could later if needed), can you start a Discussion on number testing flavors so we can stare at them and reason about what we want moving forward?

…

________________________________ From: Nick Johnson ***@***.***> Sent: Saturday, March 4, 2023 1:53:52 PM To: sandialabs/pyttb ***@***.***> Cc: Dunlavy, Daniel M ***@***.***>; Author ***@***.***> Subject: [EXTERNAL] Re: [sandialabs/pyttb] implement hosvd (Issue #56) Sounds good. I'll hold off then. I was exploring going in the directions of numbers.Real since it seemed to be a more general solution for detecting generic scalars, but unfortunately it doesn't seem to be compliant with mypy type validation. Currently across pyttb we test for numeric scalars in a variety of different ways. Eventually we probably just centralize some of that to a utility to make sure we stay consistent. — Reply to this email directly, view it on GitHub<https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fsandialabs%2Fpyttb%2Fissues%2F56%23issuecomment-1454871910&data=05%7C01%7Cdmdunla%40sandia.gov%7C182173c69bc74e1a997f08db1cf291a3%7C7ccb5a20a303498cb0c129007381b574%7C1%7C0%7C638135600379650815%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=Z1s%2B4vt1iPZSlB2Pcc6N6K9dzhmFo6U9bqk1WItJwpo%3D&reserved=0>, or unsubscribe<https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAHIY2IQTVAPV7UC4PQRPKELW2OTWBANCNFSM6AAAAAAVEOHJDY&data=05%7C01%7Cdmdunla%40sandia.gov%7C182173c69bc74e1a997f08db1cf291a3%7C7ccb5a20a303498cb0c129007381b574%7C1%7C0%7C638135600379650815%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=4RSUjlpOPlOYY%2FIfGarhf%2BlVIzd4hwkmG8FWvl8CdCM%3D&reserved=0>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[ntjohnson1]

Resolved with #67