The implementation of DSPD is wrong

[zaikunzhang]

https://github.com/blockwise-direct-search/bds/blob/893667e4590d91c068225a98dfabc862c01705d9/src/bds.m#L279-L290

This part is (still) wrong.

@OptHuang You should know this algorithm well enough to implement it correctly.

[OptHuang]

Dear Prof. Zhang,

Thanks for your reminder. First, now I know what is wrong after reading Haitian's comments and I will tell Haitian how to modify it. Second, today we two only checked and corrected comments in "src" for 3.5 hrs. Thank you again.

Best regards, Cunxin

[zaikunzhang]

First things first: MATLAB does not support loops (just imagine it in that way). Any loop is wrong. A MATLAB programmer should always keep this in mind. This applies to any scripting language (check what this means), including Python. Understand?

However, this is not the major mistake of the current implementation.

[OptHuang]

Yes, you are right. Noted with thanks.

First things first: MATLAB does not support loops (just imagine it in that way). Any loop is wrong. A MATLAB programmer should always keep this in mind. This applies to any scripting language (check what this means), including Python. Understand?

However, this is not the major mistake of the current implementation.

[Lht97]

Thanks to Dr. Zhang and Cunxin, I have corrected the way of obtaining the searching set in DSPD correctly by https://github.com/blockwise-direct-search/bds/blob/e7bf674ecda6790e7051808840688cb03f8306c9/tests/competitors/dspd.m#L224

[Lht97]

I have implemented DSPD separately and correctly under the discussion of Cunxin and Dr. Zhang.