Comments (6)
All of the models currently built-in to Ax do assume known observation noise SEM. To infer the noise level, you'll have to plug-in a separate Botorch model. You can do this by following the directions here: https://botorch.org/tutorials/custom_botorch_model_in_ax
In that tutorial, instead of defining SimpleCustomGP
you could replace that with
from botorch.models.gp_regression import SingleTaskGP
and then everywhere in the tutorial where it uses SimpleCustomGP
, just put in SingleTaskGP
instead.
If you follow all of the steps in that tutorial, you will be using a model that infers the noise level. You can then return whatever you want as the SEM in the evaluation function (0 for instance), and it will be ignored by the model.
(The only alternative to this that I can think of would be to do a few function evaluations at a single point to try and estimate the noise level directly and then use that as SEM).
from ax.
Ah, thank you.
from ax.
This is something that is in-progress, so I'll leave this issue open until it is resolved.
from ax.
There is now a fix for this on the latest stable version (0.1.5); if you just do not pass SEM, it will no longer be assumed to be 0. @72496385, which API were you using for your optimization? Let me know if you'd like me to show an example of what it would look like to just not provide SEM to it.
from ax.
Closing this issue, feel free to reopen if help is needed with how to not provide the SEM!
from ax.
Thanks a bunch for this feature.
feel free to reopen if help is needed with how to not provide the SEM!
Just a note that it took me a while to figure out that I needed to explicitly pass None
as the SEM. (I first tried passing the evaluation result alone, then the evaluation result alone in a tuple.)
from ax.
Related Issues (20)
- Issue when starting an AxClient with out-of-design points HOT 2
- cannot import name 'TrainingData' HOT 2
- applying complex constrains HOT 2
- Ax is not not starting as many workers as I'd like to; sometimes, get_next_trials returns 0 new trials HOT 5
- Evaluating custom candidates HOT 2
- Input Feature Selection - Does the relevant code exist? HOT 6
- [Feature Request] support constraints on `ChoiceParameters` HOT 4
- Extending Models.THOMPSON with an extra parameter HOT 1
- There are some questions when i use the Ax HOT 7
- Space characters in the objective name AND specifying a threshold leads to an error message: "AssertionError: Outcome constraint should be of form `metric_name >= x" HOT 2
- Pandas deprecation warning when deserializing AxClient JSON HOT 2
- AX seems to get stuck with Ray
- `StandardizeY` transform requires non-empty data." when using SAASBO
- Plotting outside of a notebook HOT 1
- Setting search space step size in Ax Service API HOT 10
- Problem when Sobol falls back to HitAndRunPolytopeSampler HOT 5
- Arms from previous batch keep appearing in new batches HOT 5
- EHVI & NEHVI break with more than 7 objectives HOT 4
- Multi-objective experiments generate duplicated data HOT 5
- Question: Transforming objective when passing `best_f` to `ProbabilityOfImprovement`, etc. HOT 3
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from ax.