Version 1.4.0#

Breaking interface changes#

Important

Starting from v1.4.0, the McmcSampler class does not use anymore labels and use_labels_dict, to select the parameters used by the sampler. Starting from this version, all the free parameters will be sampled. To select sampled parameters, please free/freeze them. Please read the caveats for the Bayesian model fitting for further details.

New features#

MCMC functionalities#

UltraNest support has been added. Please read the caveats for the Bayesian model fitting, for further details. Complete examples are available in:
- Model fitting 2: Only SSC, MCMC without minimizer step and ultranest backend with Open MPI
- Model fitting 2: Only SSC, MCMC without minimizer step and ultranest backend
A complete example for the emcee backend is available in the MCMC section of Model fitting 1: Only SSC and extended radio jet.
UltraNest and emcee share the same interface, except a few differences. Please read the caveats for the Bayesian model fitting for further details.

The McmcSampler.plot_model() method, by default, will plot the mcmc best-fit model. To plot the frequentist best-fit model, use (plot_mcmc_best_fit_model==False)

# NOTE: plot_mcmc_best_fit_model=True gets the parameters corresponding to the max log-like
p=mcmc.plot_model(sed_data=sed_data,fit_range=[1E11, 2E27],size=1000,quantiles=[0.01,0.99],plot_mcmc_best_fit_model=False)
p.setlim(y_min=1E-14,x_min=1E6,x_max=2E28)

The mcmc best-fit model can be easily restored:
```
mcmc.reset_to_mcmc_best_fit()
```
The method:meth:.McmcSampler.set_bounds now can be also used for individual parameters:
```
mcmc.set_bounds(comp_name='jet_leptonic',par_name='N',par_bounds=[1E-5,10])
```
After running the mcmc sample, the input burnin value can be updated by using the autocorrelation time of emcee. This method, is not implemented and is not needed for UltraNest.
```
mcmc.tune_burnin()
```

Leptonic equilibrium#

Added support for leptonic equilibrium jet models using injected emitters distributions. See Leptonic equilibrium

The model can be created by passing an InjEmittersFactory distribution to Jet:

from jetset.jet_model import Jet
from jetset.jet_emitters_factory import InjEmittersFactory

q_inj = InjEmittersFactory().create_inj_emitters('pl')
my_jet = Jet(emitters_distribution=q_inj)

Validation against the Fokker-Planck time-dependent solution in JetTimeEvol: Validation of the leptonic equilibrium against the Fokker-Plank equation solution
End-to-end model fitting workflow with MCMC for leptonic equilibrium models: Model fitting: Leptonic equilibrium SSC

Radio component in composite models#

Added a new analytical radio component, RadioSpectrum, for extended jet radio emission with self-absorption turnover and high-frequency cutoff. See Jet Radio Component: RadioSpectrum

RadioSpectrum can be combined with a jet model in FitModel:

from jetset.jet_radio_component import RadioSpectrum
from jetset.model_manager import FitModel

fit_model = FitModel(jet=prefit_jet, name='SSC-best-fit-lsb', template=None)
fit_model.add_component(RadioSpectrum())
fit_model.composite_expr = '(jet_leptonic+radio_spectrum)'

A complete fitting workflow (SSC + extended radio component, minuit and MCMC) is available in Model fitting 1: Only SSC and extended radio jet

Corona#

Added EC support for Corona, see: Example for using a Corona within the EC scenario.

data handling#

UL are handled when rebinning SED data
added nu min/max to rebin
ASI ssdc sedbuilder

Pair production for external fields#

added pair production absorption for Dusty Torus, Broad Line Region and Corona.
internal absorption can be enabled for EC components (DT/BLR) and used in fitting workflows. See Internal absorption and Model fitting 3: External Compton
optional use_R_H_profile_extrapolation in internal absorption setup to speed up evaluations:
```
jetset_model.enable_internal_absorption('DT', use_R_H_profile_extrapolation=True)
```
A complete fitting workflow with EC and internal absorption see: Model fitting 3: External Compton

bug fixing#

minor bug fixes
fixed typos in variable names
fixed a bug in serialization of depending parameter.

C codebase#

Substantial refactoring of the C code, dividing the blob structure in several sub-structures, with hierarchy.

installation#

Added source installation via ./install.sh.
Detects if you are in a conda/mamba/micromamba evn or a pip evn, and install JetSeT from source, together with the dependencies
When run inside an active conda/mamba environment, ./install.sh syncs JetSeT constraints from requirements.txt into $CONDA_PREFIX/conda-meta/pinned. This pins JetSeT requirements in that environment and helps prevent later conda/mamba installs of other packages from overriding them. For pip virtual env, the problem
This script is recommended when binaries are not available for your Python/platform, or for developers and advanced users.
Run the script from the project root to install dependencies, compile extensions, and install JetSeT:
```
./install.sh
```
If you want to skip the installation of dependencies
```
./install.sh -skip-dep
```
If you want to experiment with different requirements (do it on your own risk):
- edit the requirements.txt accordingly and:
```
./install.sh
```

Verify the installation with, optional (not within the code repo dir):

python -c "import jetset; print(jetset.__version__)"
pytest --disable-warnings --pyargs -vvv jetset.tests.test_users::TestUser

Full source-install requirements and troubleshooting are documented in Installation (including SWIG notes in Swig installation).