Comments (4)
For NNMF_multiplicative
:
Based on the existing code, I'm assuming the threshold in question is fit_residual
falling below fit_error_limit
. 128c7bf implements this.
For NNMF_sgd
:
Currently, nothing analogous to fit_error_limit
exists, although the docstring suggests that it should. That's easy to add, but (assuming this is the goal), we'd also have to add a step that computes something analogous to fit_residual
. I think that requires calling predict()
after each iteration, which would have a performance cost.
Do you think it's worth implementing this for NNMF_sgd
? Happy to do it if so, just wanted to check if it makes sense.
from neighbors.
I think the pattern that might make more sense is to loop over iterations and then break of residual is lower than the tolerance.
Check out some of the sklearn algorithms, such as NNMF https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/decomposition/nmf.py
from neighbors.
Sure, can definitely do that. It probably also makes sense to check every 10 iterations, as sklearn does, instead of each time.
One question on the math: sklearn is calculating error by comparing current error to the initial error from before fitting began. With multiplicative updating, (previous error - current error) / initial error
(link), and with coordinate descent, current error / initial error
(link).
Should we do the same?
from neighbors.
Here are some quick thoughts
- Let's go with percent error threshold
- Let's add a rate of change threshold
- let's try to match Sci-kit learn's defaults in terms of order of operations.
from neighbors.
Related Issues (20)
- Double check code for NNMF_sgd since refactor. HOT 1
- Consider adding NNMF_als class
- Add Tensor Extension HOT 1
- Correct sorting issue with create_sub_by_item_matrix
- Might consider adding an automask if missing values/sparsity is detected HOT 1
- autoscaling of colorbar for plot_predictions
- Documentation
- Documentation
- Add support for sparse matrices
- Speed up SGD with Parallelization HOT 1
- BUG - dilating timeseries converts nans to zeros
- README intro missing import statement
- Question about estimate_performance() HOT 1
- estimate_performance sometimes fails when return_agg is True
- unflatten_dataframe produces warnings on repeated calls, e.g. with estimate_performance
- Add ability to pass any kernel
- Convolution position.
- Prevent NaN propagation in SGD
- Move docs to jupyter book
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 neighbors.