Skip to content

Latest commit

 

History

History
39 lines (31 loc) · 1.64 KB

coming_soon.md

File metadata and controls

39 lines (31 loc) · 1.64 KB
layout permalink title
home
/coming_soon/
Seldonian \| Coming soon
Coming soon
<h5>Major features planned for Spring 2023 release, in order of priority:</h5>
<ul>
    <li>
    <a href="https://en.wikipedia.org/wiki/Hoeffding%27s_inequality">Hoeffding's</a> concentration inequality bounding method. This will enable running true Seldonian algorithms (as opposed to quasi-Seldonian algorithms) with the Seldonian Engine.
    </li>
     <li>
    Multiclass classification - This was implemented in the Engine in <a href="https://pypi.org/project/seldonian-experiments/0.0.8/">version 0.0.8</a>, but is not yet integrated into the Experiments library.
    </li>

    <li>
    Multiple label columns in a dataset (supervised learning).
    </li>

    <li>
    Importance sampling variants, such as weighted and per-decision importance sampling. In version alpha, the standard importance sampling estimator is the only primary objective function available for Seldonian reinforcement algorithms. 
    </li>

    <li>
    An automated method for determining optimal data split between candidate data and safety data.
    </li>

    <li>
    Automatic differentiation using <a href="https://github.com/google/jax">JAX</a>. JAX will be a big upgrade to autograd and should provide significant improvements for many problems. Autograd will remain an option for users who do not wish to install or use JAX. 
    </li>
    
</ul>