Farnazmdi commented 2 years ago

To have equal transitions between the states (and not estimating the T), in BaumWelch.py, during the M step, the transition probabilities are set manually according to the number of states. (only one line is added, and line 89 is commented to prevent the T estimation). Other changes are some clean ups.
The function that is used for performing the cross-validation and finding the likelihood of masked observations (output_LL) is moved to the crossval.py
figure 19 is the figure where we perform cross validation on synthetic data.
figure 21 is the figure we perform cross validation on experimental data.

codecov[bot] commented 2 years ago

Codecov Report

Merging #957 (9dd5c0a) into master (ba1e50d) will decrease coverage by 0.41%. The diff coverage is 62.06%.

:exclamation: Current head 9dd5c0a differs from pull request most recent head 335a2cb. Consider uploading reports for the commit 335a2cb to get more accurate results

@@            Coverage Diff             @@
##           master     #957      +/-   ##
==========================================
- Coverage   75.46%   75.05%   -0.41%     
==========================================
  Files          26       26              
  Lines        2164     2169       +5     
==========================================
- Hits         1633     1628       -5     
- Misses        531      541      +10

Flag	Coverage Δ
unittests	`75.05% <62.06%> (-0.41%)`	:arrow_down:

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
lineage/BaumWelch.py	`89.24% <ø> (ø)`
lineage/plotTree.py	`0.00% <0.00%> (ø)`
lineage/crossval.py	`82.50% <68.18%> (-17.50%)`	:arrow_down:
lineage/states/stateCommon.py	`100.00% <100.00%> (ø)`
lineage/tests/test_CrossVal.py	`100.00% <100.00%> (ø)`
lineage/Analyze.py	`92.78% <0.00%> (-1.04%)`	:arrow_down:

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more

aarmey commented 1 year ago

@Farnazmdi is this something that should eventually be merged or deleted?

Farnazmdi commented 1 year ago

There is an underflow error here in calculating the gamma likelihood when building figure 21 (cross validation for experimental data of lapatinib and gemcittabine): outt -= np.sum(gammas[i][jidx] * np.log(gamP)) FloatingPointError: underflow encountered in multiply in gamma_LL line 65.

@aarmey Do you have an idea about how to address this? It is already in log format..

aarmey commented 1 year ago

I made a small adjustment that I think will fix the issue. Let me know if it still errors.

Farnazmdi commented 1 year ago

Thank you, it is fixed.

Farnazmdi commented 1 year ago

This should be ready to merge.

aarmey commented 1 year ago

Great. Merging.

meyer-lab / tHMM

Cross validation with equal transitions #957

Codecov Report