ahwillia

from pca_crossval import * 
from tqdm import tqdm
import itertools
np.random.seed(1111)

m, n, r = 100, 101, 3
Utrue, Vtrue = rand(m,r), rand(r,n)
data = np.dot(Utrue, Vtrue) + .25*randn(m,n)
data[data < 0] = 0
ranks = []

ahwillia

import numpy as np
import scipy.linalg

def elastic_net(A, B, x=None, l1=1, l2=1, lam=1, tol=1e-6, maxiter=10000):
    """Performs elastic net regression by ADMM

    minimize ||A*x - B|| + l1*|x| + l2*||x||

    Args:
      A (ndarray) : m x n matrix

ahwillia

def reverse_segment(path, n1, n2):
    """Reverse the nodes between n1 and n2.
    """
    q = path.copy()
    if n2 > n1:
        q[n1:(n2+1)] = path[n1:(n2+1)][::-1]
        return q
    else:
        seg = np.hstack((path[n1:], path[:(n2+1)]))[::-1]
        brk = len(q) - n1

ahwillia

import numpy as np
from numpy.random import randn, rand
from scipy.optimize import minimize
import matplotlib.pyplot as plt
from nnls import nnlsm_blockpivot as nnlstsq
import itertools
from scipy.spatial.distance import cdist

def censored_lstsq(A, B, M):
    """Solves least squares problem with missing data in B

ahwillia

import numpy as np
from tqdm import trange
import matplotlib.pyplot as plt


# TODO: subclass np.ndarray?
class ShiftMatrix(object):
    """
    Thin wrapper around a numpy matrix to support shifting along the second
    axis and padding with zeros.

ahwillia

"""
A simple implementation of Poisson regression.
"""

import numpy as np
from scipy.optimize import minimize

n = 1000   # number of datapoints
p = 5     # number of features

ahwillia

"""
Computing hessian-vector products in tensorflow.

For simplicity, we demonstrate the idea on a Poisson regression model.
"""

import tensorflow as tf
import numpy as np
from scipy.optimize import minimize

ahwillia

def bandpass(x, lowcut, highcut, fs, order=5, axis=-1, kind='butter'):
    """
    Parameters
    ----------
    x : ndarray
        1d time series data
    lowcut : float
        Defines lower frequency cutoff (e.g. in Hz)
    highcut : float
        Defines upper frequency cutoff (e.g. in Hz)

ahwillia

import numpy as np
from scipy.linalg import solve_circulant, circulant
from numpy.testing import assert_array_almost_equal
import numba


@numba.jit(nopython=True, cache=True)
def rojo_method(c, a, f, x, z):
    """
    Solves symmetric, tridiagonal circulant system, assuming diagonal

ahwillia

"""
References:

  - B. Plateau, On the stochastic structure of parallelism and synchronization models for distributed algorithms.
  Perform. Eval. Rev., 13 (1985), pp. 147–154.

  - Dayar, T., & Orhan, M. C. (2015). On vector-Kronecker product multiplication with rectangular factors.
  SIAM Journal on Scientific Computing, 37(5), S526-S543.
"""