vmboehm / mnulfi

throwing LFI on massiveNus sims

MnuLFI

Testing strengths and pitfalls of likelihood-free inference with neural networks on the MassiveNus peak count data set [1,2]

Installing the MnuLFI modules

Clone the repo then run

pip install --user -e .

Then to test the install and generate the data simply run

pytest

mnulfi's Issues

non-zero covariance mixture models

In the current pyDELFI implementation, mixture models for NDE can have zero covariance, which results in weird posteriors. @VMBoehm @EiffL

IMNN

implement IMNN data compression for MassiveNuS peak counts

Regularization

Implement regularization for the NDE training.

data compression validation

Compare using Gaussian pseudo-likelihood the posterior derived from score, PCA, etc compressed data vectors to the posterior from the the peak count data vector to validate the different compression methods.

UT not defined

Traceback (most recent call last):
File "", line 1, in
File "/home/nessa/Documents/codes/MnuLFI/mnulfi/data.py", line 50, in PeakCnts_LHC
peakct = np.load(os.path.join(UT.dat_dir(), 'avg_peakcnts_lhc.npy'), allow_pickle=True)
NameError: name 'UT' is not defined

Running Data._make_data()

When I try to run Data._make_data(), it is giving me this error

Traceback (most recent call last):
File "", line 1, in
File "/Users/arinavsar/Desktop/MnuLFI-master/mnulfi/data.py", line 130, in _make_data
avg_peak = np.average(peaks, axis=1)
File "<array_function internals>", line 6, in average
File "/Users/arinavsar/anaconda3/envs/mnulfi/lib/python3.7/site-packages/numpy/lib/function_base.py", line 390, in average
avg = a.mean(axis)
File "/Users/arinavsar/anaconda3/envs/mnulfi/lib/python3.7/site-packages/numpy/core/_methods.py", line 138, in _mean
rcount = _count_reduce_items(arr, axis)
File "/Users/arinavsar/anaconda3/envs/mnulfi/lib/python3.7/site-packages/numpy/core/_methods.py", line 57, in _count_reduce_items
items *= arr.shape[ax]

cloud versus skewers

• How the skewer set up of MassiveNuS impacts NDE?
• If we convolve the skewers, does this improve NDE posteriors?

NDE adaptive learning rate

Implement adaptive learning rate for the pyDELFI NDE training @VMBoehm @EiffL