fix Julia 1.7 tests #85
Conversation
|
I'm a little bit confused as to why this improves type-stability. Do we have a global |
|
Actually, I think the |
|
Test failure is due to optimisation having ended up in a bad local optimum ... |
st--
changed the title
st--
changed the title
Hmmm that's weird. Shouldn't the optimisation problem be convex, since the likelihood is log-concave? |
The inner loop should be convex optimisation, but not the optimisation of hyperparameters using the approximate log marginal likelihood.. But yeah I don't understand why that would have changed so much in the 1.6 -> 1.7 change :S |
|
Might be because they changed the default RNG? |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Codecov Report
@@ Coverage Diff @@
## master #85 +/- ##
==========================================
- Coverage 90.78% 90.13% -0.65%
==========================================
Files 4 4
Lines 282 294 +12
==========================================
+ Hits 256 265 +9
- Misses 26 29 +3
Continue to review full report at Codecov.
|
|
I'm still a bit confused by the need for |
| # The random number generator changed in 1.6->1.7. The following vector was generated in Julia 1.6. | ||
| # The generating code below is only kept for illustrative purposes. | ||
| #! format: off | ||
| Y = [0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0] |
There was a problem hiding this comment.
Couldn't we just use rng=MersenneTwister() or whatever?
|
@willtebbutt It's just a nicer way of initialising the variable no? |
🤦 you're right. |
No description provided.