Comments (4)
Moreover, I just tested !pip install git+https://github.com/mirt001/bfast.git@mirt001-testing-k-1-2#egg=bfast
branch on my fork, which comments the lines that check whether k is in k_valid.
It works with seemingly correct results on google colab, and I assume nothing exploded on google side of things.
This is far from proper testing, but whatever prompted imposing the k>2 limitation, might not be in the code anymore.
from bfast.
This limitation has been there from before I joined the project, so I don't know exactly why it is there. I think it has something to do with performance of the GPU version. 2k + 1 is the inner dimension of multiple vector/matrix operations in the code, and the GPU kernels are tuned with this assumption in mind. It should still run for k=1 and k=2, but the performance would probably be suboptimal. More testing needs to be done.
from bfast.
@mortvest I understand that performance might be suboptimal, when compared to k=3, but the performance is still better than the R bfast. Also, python and opencl backends are at least comparable with k=1, and I believe opencl is still faster.
At most, the performance decrease for k=1 and k=2 for opencl should be documented and left to the user. Especially since, imo, most users need specifically k=1 and k=2.
What do you have in mind for testing? I have to run bfastmonitor quite a few times these days. If I can watch out for something extra that would help with development, that would be great. I am currently running my fork that allows me to run with k=1 and k=2. No issues so far.
from bfast.
Thanks for your input @mirt001, I didn't know that it was a popular setup. By testing, I meant how much performance decrease there is and if it actually produces the correct results. The latter is probably true. Regarding the former, we would probably need to retune some parameters for the GPU version. I'll look into it next week.
from bfast.
Related Issues (20)
- Error running BFASTMonitor example on a CPU: Build Program Failure in one env and kernel dies in another HOT 2
- import bug fix in the documentation HOT 2
- The error message of level does not reflect its actual behaviour HOT 1
- performance benchmark between the different backend ? HOT 11
- drop usage of sklearn ? HOT 2
- pyhton backend n_chunk is never used HOT 3
- why _model is reset each time fit is called ? HOT 3
- OpenCL backend is way too slow compared to futhark bench HOT 1
- Write an example that demonstrates usage of multiple GPUs
- GPU data transfer overhead amortization
- BFAST cannot be install from git source
- Version 0.7 fails to install on google colab HOT 2
- use bfast for one dimentional time series data HOT 10
- numpy should not be referenced in the setup.py HOT 1
- Stable period determination not implemented HOT 2
- How to get full list of breaks for one pixel HOT 2
- Error in normal terminal and Google colab with Get Started example HOT 1
- IndexError when datetime is used with hour != 0
- Support for multiple detected breaks
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 bfast.