andres-fr

#!/usr/bin python
# -*- coding:utf-8 -*-


"""
"""


import os
import h5py

andres-fr

import matplotlib.pyplot as plt
import matplotlib.font_manager as plt_fm

class PltFontManager:
    """
    Sometimes matplotlib finds the system font paths, but setting them can
    still be challenging due to using the wrong name or matplotlib complaining
    about missing elements.

    This manager static class is intended to aid with that, by facilitating

andres-fr

from collections import defaultdict
import random

class SubsetSampler(torch.utils.data.Sampler):
    """
    Like ``torch.utils.data.SubsetRandomsampler``, but without the random,
    and with the possibility of balanced sampling. Samples a subset of the
    given dataset from a given list of indices, without replacement.
    Usage example::

andres-fr

#!/usr/bin/env python
# -*-coding:utf-8-*-


"""
This module contains 2 functions:

* ``cose_recovery``: Implements the L1 compressed sensing program, allowing
  for L2 error tolerance in the constraint.
* ``cose_recovery_mpp``: Parallelized version of ``cose_recovery`` that runs

andres-fr

#!/usr/bin/env python
# -*- coding:utf-8 -*-


"""
!!! WARNING !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

The quantities computed in this gist are wrong. While a correction is
pending, see the following discussion for more details. The TLDR is that
they must be divided by batch_size^2, or use the alternative Cockpit

andres-fr

#!/usr/bin/env python
# -*-coding:utf-8-*-


"""
This module contains a minimal example to illustrate one of the main
implications of the compressed sensing paradigm: That a perfect recovery of a
sparse signal is possible with a number of measurements that depends on the
signal sparsity, and not on the signal length.

andres-fr

interactive_gpu()
{
    # needed by getopt: https://stackoverflow.com/a/5048356
    local OPTIND

    # default parameter values
    seconds=300
    partition="large"

    # parse flags to optionally override param defaults

andres-fr

#!/usr/bin/env python
# -*- coding:utf-8 -*-


"""
Most existing breadth-first search implementations provide a solution to
*traverse* a tree, but they don't *return* that specific tree. This module
provides a tree datastructure and BFS implementation to obtain the BFS tree.
"""

andres-fr

#!/usr/bin/env python
# -*- coding:utf-8 -*-


"""
"""


import librosa
import numpy as np

andres-fr

import jax.numpy as jnp
from jax import jit, value_and_grad
from jax.config import config
config.update("jax_debug_nans", True)

class RegularInv1dSampler:
    """
    This regular inverse sampler deals with the following problem: given a
    smooth function ``y=f(x)`` for ``x`` scalar and ``y`` n-dimensional,
    retrieve ``N`` monotonically increasing values of ``x``, such that the