T Ehrhardt
T Ehrhardt
Several services implement numpy generalised universal functions using `from numba import guvectorize`. Unless I'm misunderstanding, `numba.set_num_threads(PISA_NUM_THREADS)` in https://github.com/icecube/pisa/blob/master/pisa/utils/numba_tools.py#L71 is always executed (for the parallel target) only after some involved chain...
This attribute claims to disable caching at the [Stage](https://github.com/icecube/pisa/blob/master/pisa/core/stage.py) base class level, but it doesn't actually seem to have any functionality. No service throughout the code base implements any custom...
Cursory literature search revealed two recent algorithms which might be worth investigating: * [NuOscProbExact](https://inspirehep.net/literature/1731803) (possibly improved upon in https://inspirehep.net/literature/2697207) * fully in Python * [NuFast](https://inspirehep.net/literature/2783495) * fast C++ and Fortran...
Blindness levels affect fit behaviour in * https://github.com/icecube/pisa/blob/dacc631a04821003aa012e74908bda689c5936c3/pisa/analysis/analysis.py#L1977 * https://github.com/icecube/pisa/blob/dacc631a04821003aa012e74908bda689c5936c3/pisa/analysis/analysis.py#L2181 * https://github.com/icecube/pisa/blob/dacc631a04821003aa012e74908bda689c5936c3/pisa/analysis/analysis.py#L2190 * https://github.com/icecube/pisa/blob/dacc631a04821003aa012e74908bda689c5936c3/pisa/analysis/analysis.py#L2360 * https://github.com/icecube/pisa/blob/dacc631a04821003aa012e74908bda689c5936c3/pisa/analysis/analysis.py#L2367
* an analyser making use of https://github.com/icecube/pisa/blob/dacc631a04821003aa012e74908bda689c5936c3/pisa/analysis/analysis.py#L963 is unable to switch off recording and storing of fit/minimiser history, best-fit asimov template, detailed metric maps * depending on nature of analysis,...
These do not implement `apply_function` and thus assume that no preceding service has the ability to reset or vary the "nominal" fluxes. [`mceq_barr_red`](https://github.com/icecube/pisa/blob/70f71538fc5eadccb3419fae4a27a24781d75cdb/pisa/stages/flux/mceq_barr_red.py) states at least that it is intended...
* general problem running in single-precision FP mode? (`pisa_tests/test_kde_stage.py` fails with [Error: lists contain a non-float value](https://github.com/icecube/kde/blob/master/kde/kde.c)) * but `pisa_tests/test_kde_stage.py` also fails to `assert np.abs(total_no_lin / total_with_lin - 1.0) <...
... in particular given how central it is.