ROADMAP.md

Longer term Roadmap

1. Dataframe is key underlying datatype (as opposed to Matrix{Float64})
2. Performance benchmarking and optimization
3. Robust sampling methods
4. Build out metrics, potentially break out as separate repo
5. Testing against more real world datasets

Short-term TODO's

1. Partial/sensitivity refactor + unit tests
2. More unit tests
   • likelihood under different observations for bart.jl
   • split.jl
3. Graphical output for experiments
4. Experiments with multiple datasets
5. Support categorical predictors
6. Track & optionally show metadata and state during training
7. Support minibatch sgd during neural net training
8. Supervised regression support for neural nets

Experiments To Run

1. Impact of the jump distribution on time to convergence
2. Convergence of MCMC results