Comments (2)
This bug seems extremely unstable. I don't observe the same behavior, but I was able to excite it with some fooling around, even isolating down to 200 examples. As far as I can tell, this is a feature encoding failure---in particular when I examine one of the offensive examples with --audit, it looks like 'verdana' comes up as a feature many times. This leads to instability, because a change in the feature's weight is multiplied many times. I'd suggest collapsing such features to one feature with a large value which allows the normalization elements of the update rule to kick in.
from vowpal_wabbit.
John, thanks so much for the insight.
Look like I have more work to do on feature generation, and maybe clean up my data-set a bit more (skipping a few of the noisier features)
from vowpal_wabbit.
Related Issues (20)
- can't get proper prediction score from python when --boosting N option used at training time HOT 2
- First regression test
- Default FTRL hyperparameters are very conservative HOT 4
- Sklearn adapter function `tovw` does not support unsigned integers in features HOT 3
- vw core CLI: --nn data-leak: SGD update is incorrect HOT 5
- Vowpal_wabbit failed to runtest on MSVC on Windows HOT 2
- Coin/FTRL/Pistol are not available via vw.get_config() HOT 1
- Detailed explanation on --explore_eval option for contextual bandits HOT 2
- Classification Multivariate time series : values both categorical and continues HOT 5
- Segmentation fault in CATS HOT 2
- Contextual Bandit vowpal_wabbit training dataset validation HOT 2
- New line is misinterpreted as example
- Binary File Inputs HOT 3
- Your domain is only being misused for illegal gambling promotions in Indonesia HOT 2
- --interact is not working HOT 3
- Slates Json parser error
- Option to enable SSE2 optimization HOT 2
- Segfault on ccb_explore_adf -cb_type dr HOT 7
- Support for multi-line featuers in --audit_regressor HOT 1
- New installation unable to find boost python lib LINK : fatal error LNK1104: cannot open file 'boost_python312-vc143-mt-x64-1_84.lib' HOT 2
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 vowpal_wabbit.