-
- rename write_gradients -> write_gradient
Even though there may be multiple elastic parameters, there is only one gradient. It is not correct to speak of multiple gradients.
There may be some unde…
-
See [here](https://github.com/libqueso/queso/blob/dev/src/core/inc/GslOptimizer.h#L75). Should be `STEEPEST_DESCENT`, not `STEEPEST_DECENT`.
-
I brought up in the comments on #352 that it would be useful to be able to restrict the application of a `StepRule` to only certain parameters.
A useful trick with convolutional nets is to scale the …
-
Following a conversation with @bartvm I now understand why `Momentum` is implemented the way it is (it's meant to be chained with `SteepestDescent`, so that `Momentum`'s `self.momentum * velocity + gr…
-
The unit tests are broken again. It appears they are failing on tests that involve functions like `calculate_steepest_descent_across_cell_faces` that now use masked arrays. @SiccarPoint, I think these…
-
At the moment, when computing the steepest descent images, we warp the gradients only in the forward mode, by passing in a `kwarg` of `forward=(self.template, self.optimal_transform, self._warp)`.
Th…
-
what do agents do when they get disconnected. They can't move towards their respective targets since they have lost connectivity
-
Dealing with bound constraints in the optimize problem.
---
- Bitbucket: https://bitbucket.org/rcockett/simpeg/issue/12
- Originally Reported By: [ ](http://bitbucket.org/rowanc1)
- Originally Create…
-
At the moment, we compute the error for each single pixel. This seems to be a waste of computation.
- GPU: A good heuristic seems to be to only use every n-th pixel (especially for high pyramid level…