Tridiminished icosahedron: line search failed

glen commented

2025-07-22 21:32:09 +00:00

Owner

To reproduce: starting from the assembly produced by selecting tridiminished tetrahedron, fill in each A-A, A-B, B-C, and C-C inversive distance at -0.25 (as prescribed by the comments in the test assembly code) in turn from top to bottom, then start filling in the A-C inversive distances at -0.6545084971874737. Line search failed for me upon adding the A_1 C_3 distance (the second such distance after A_1 C_2.

This could use independent reproducing to see if it is consistent across machines. Once replicated, it should be labeled "bug".

To reproduce: starting from the assembly produced by selecting tridiminished tetrahedron, fill in each A-A, A-B, B-C, and C-C inversive distance at -0.25 (as prescribed by the comments in the test assembly code) in turn from top to bottom, then start filling in the A-C inversive distances at -0.6545084971874737. Line search failed for me upon adding the A_1 C_3 distance (the second such distance after A_1 C_2. This could use independent reproducing to see if it is consistent across machines. Once replicated, it should be labeled "bug".

Vectornaut commented

2025-07-22 21:37:46 +00:00

Member

fill in each A-B, B-B, B-C, and C-C inversive distance at -0.25

Trying to reproduce now. There are no B–B inversive distance regulators; did you mean A–A?

> fill in each A-B, B-B, B-C, and C-C inversive distance at -0.25 Trying to reproduce now. There are no B–B inversive distance regulators; did you mean A–A?

glen commented

2025-07-22 21:41:16 +00:00

Author

Owner

Yes typo fixed now.

Vectornaut commented

2025-07-22 22:03:01 +00:00

Member

Huh: I don't get a line search failure. In fact, once I've set all the inversive distances except the A–C ones, I find that the A–C ones already have the desired values. This is consistent with our conjectural yes answer to the tridiminished icosahedron rigidity problem. Could you post the values you get for some of the A–C inversive distances once you've set all the others?

Huh: I don't get a line search failure. In fact, once I've set all the inversive distances except the A–C ones, I find that the A–C ones already have the desired values. This is consistent with our conjectural *yes* answer to the [tridiminished icosahedron rigidity](wiki/Test-problems#tridiminished-icosahedron-rigidity) problem. Could you post the values you get for some of the A–C inversive distances once you've set all the others?

glen commented

2025-07-22 22:12:18 +00:00

Author

Owner

OK will try again post-merge.

Vectornaut commented

2025-07-22 22:12:56 +00:00

Member

To help check whether we're following the same procedure: I'm setting distances in this order. The ordering of the elements in the To column might be machine-dependent, because the regulator lists in the outline view are sorted only by number of subjects, leaving the order otherwise arbitrary. (Maybe it would be useful to do a secondary sorting by subject identifiers, to match the current hard-coded outline order—but we're still planning to replace the outline view anyway.)

From	To
`A_1`	`B_2`
"	`B_3`
"	`A_3`
"	`A_2`
`A_2`	`A_3`
"	`B_3`
"	`B_1`
`A_3`	`B_1`
"	`B_2`
`B_1`	`C_1`
`B_2`	`C_2`
`B_3`	`C_3`
`C_1`	`C_3`
"	`C_2`
`C_2`	`C_3`

To help check whether we're following the same procedure: I'm setting distances in this order. The ordering of the elements in the *To* column might be machine-dependent, because the regulator lists in the outline view are sorted only by number of subjects, leaving the order otherwise arbitrary. (Maybe it would be useful to do a secondary sorting by subject identifiers, to match the current hard-coded outline order—but we're still planning to replace the outline view anyway.) From | To --- | --- $A_1$ | $B_2$ " | $B_3$ " | $A_3$ " | $A_2$ $A_2$ | $A_3$ " | $B_3$ " | $B_1$ $A_3$ | $B_1$ " | $B_2$ $B_1$ | $C_1$ $B_2$ | $C_2$ $B_3$ | $C_3$ $C_1$ | $C_3$ " | $C_2$ $C_2$ | $C_3$

glen commented

2025-07-22 22:25:46 +00:00

Author

Owner

OK, worked this time. Perhaps the first time I accidentally put a -0.25 into one of the A-C boxes, which would of course ultimately create an impossible configuration.

glen closed this issue

2025-07-22 22:25:46 +00:00

Vectornaut commented

2025-07-22 22:29:42 +00:00

Member

@glen wrote in #104 (comment):

OK, worked this time. Perhaps the first time I accidentally put a -0.25 into one of the A-C boxes, which would of course ultimately create an impossible configuration.

Yes, I recall experiencing some realization failures like that when I was playing with this assembly! I think it's an encouraging example of dyna3 working as we'd hoped (although it also highlights a pain point in the interface).

@glen wrote in https://code.studioinfinity.org/StudioInfinity/dyna3/issues/104#issuecomment-3093: > OK, worked this time. Perhaps the first time I accidentally put a -0.25 into one of the A-C boxes, which would of course ultimately create an impossible configuration. Yes, I recall experiencing some realization failures like that when I was playing with this assembly! I think it's an encouraging example of dyna3 working as we'd hoped (although it also highlights a pain point in the interface).

glen commented

2025-07-22 22:37:22 +00:00

Author

Owner

I mean neither of this thinks this is a usable app for actual modeling. I may be doing an impossible solid build at CSU-Pueblo in September. If that is confirmed, I would like to try dyna3 with it displaying its best fit upon realization failure to see if I can (a) reproduce orchidpalms' result on its nearest-miss dome over a 5-5-4 vertex and in fact (b) see if the quality-of-fit improves by replacing the rigid octagon on top by a dome of equitriangles (so there are only triangles besides the initial vertex). In that case, we would totally pump up the priority of (I) being able to display the best fit even when realization fails, and (II) having easier ways to set segment lengths and/or manipulate the display.

Rows
Columns

Tridiminished icosahedron: line search failed #104