Closed jonathanishhorowicz closed 5 years ago
Merging #37 into master will decrease coverage by
1.45%
. The diff coverage is85.71%
.
@@ Coverage Diff @@
## master #37 +/- ##
==========================================
- Coverage 77.09% 75.63% -1.46%
==========================================
Files 20 20
Lines 1004 1022 +18
==========================================
- Hits 774 773 -1
- Misses 230 249 +19
Impacted Files | Coverage Δ | |
---|---|---|
src/abc/simulation.jl | 100% <ø> (ø) |
:arrow_up: |
src/abc/io.jl | 17.44% <ø> (ø) |
:arrow_up: |
src/abc/model_selection_io.jl | 100% <100%> (ø) |
:arrow_up: |
src/abc/emulation.jl | 100% <100%> (ø) |
:arrow_up: |
src/abc/rejection.jl | 90.32% <82.05%> (-6.35%) |
:arrow_down: |
src/abc/smc.jl | 86.06% <83.75%> (-7.66%) |
:arrow_down: |
src/abc/model_selection.jl | 86.44% <87.59%> (+0.89%) |
:arrow_up: |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update e034f24...713c720. Read the comment docs.
Put sections of parameter code into function (e.g.
check_particle
andcheck_particle_batch
to compare particles to data. These functions are then called by model selection codeThis meant I could remove the arguments
for_model_selection
andnormalise_weights
from the ABC functionsBatch size has been removed from the SMC tracker and input and is now an argument of these new functions. This is so that the model selection code can call it without creating a new input or tracker at each iteration (the batch sizes changes at each model selection iteration).
Model selection print statements updated to use
info()
Model selection retraining not currently being used due to PosDefException - this may be because the retraining code is not perturbing samples from the previous population's posterior
Deleted Untitled.ipynb
Changed the names of the user-level model selection functions to
SimulatedModelSelection
andEmulatedModelSelection