-
Notifications
You must be signed in to change notification settings - Fork 1.9k
DSJSON
The DJSON format is geared towards expressing Contextual Bandits problems (CB, CCB and Slates currently).
All elements at the top level are problem specific properties. In particular:
Version
- must be the string 1
and indicates the version of the format to use.
c
- container of all examples.
Inside c
, there are a few special fields that you can define:
_multi
- an array of examples with one per action.
_slots
- an array of example with one per slot, to be used with problems that requires those such as CCB and Slates.
All other fields within "c" are considered part of the shared example.
Features are JSON strings, integer, float, boolean, arrays of integers and/or floats. Namespaces are defined using objects where the namespace is the field name.
The timestamp field Timestamp
should follow the following format: year-month-dayThh:mm:ss.ssssZ
. For example: 2021-07-29T10:00:00.999Z
Most produces will trim timestamps on the second, leading to milliseconds always being zero.
Example of CB payload:
{
"Timestamp": "1999-10-10T13:20:20.500Z",
"Version": "1",
"c": {
"TShared": {
"a": 1,
"b": "x"
},
"c": 10,
"_multi": [
{
"TAction": {
"value=0.000000": 1
}
},
{
"TAction": {
"value=1.000000": 1
}
},
],
},
"VWState": {
"m": "N/A"
}
}
- Home
- First Steps
- Input
- Command line arguments
- Model saving and loading
- Controlling VW's output
- Audit
- Algorithm details
- Awesome Vowpal Wabbit
- Learning algorithm
- Learning to Search subsystem
- Loss functions
- What is a learner?
- Docker image
- Model merging
- Evaluation of exploration algorithms
- Reductions
- Contextual Bandit algorithms
- Contextual Bandit Exploration with SquareCB
- Contextual Bandit Zeroth Order Optimization
- Conditional Contextual Bandit
- Slates
- CATS, CATS-pdf for Continuous Actions
- Automl
- Epsilon Decay
- Warm starting contextual bandits
- Efficient Second Order Online Learning
- Latent Dirichlet Allocation
- VW Reductions Workflows
- Interaction Grounded Learning
- CB with Large Action Spaces
- CB with Graph Feedback
- FreeGrad
- Marginal
- Active Learning
- Eigen Memory Trees (EMT)
- Element-wise interaction
- Bindings
-
Examples
- Logged Contextual Bandit example
- One Against All (oaa) multi class example
- Weighted All Pairs (wap) multi class example
- Cost Sensitive One Against All (csoaa) multi class example
- Multiclass classification
- Error Correcting Tournament (ect) multi class example
- Malicious URL example
- Daemon example
- Matrix factorization example
- Rcv1 example
- Truncated gradient descent example
- Scripts
- Implement your own joint prediction model
- Predicting probabilities
- murmur2 vs murmur3
- Weight vector
- Matching Label and Prediction Types Between Reductions
- Zhen's Presentation Slides on enhancements to vw
- EZExample Archive
- Design Documents
- Contribute: