apatil

# Illustrates why log-sums are better than direct sums for probability integrals
# when probabilities are initially computed on the log scale

import pymc
import numpy 

lp_big = numpy.random.normal(loc=0,scale=2,size=1000)
print numpy.log(numpy.mean(numpy.exp(lp_big)))
print pymc.flib.logsum(lp_big)-numpy.log(len(lp_big))

apatil

import numpy as np
import pylab as pl
import pymc as pm

Tau_test = np.matrix([[ 209.47883244,   10.88057915,   13.80581557],
                       [  10.88057915,  213.58694978,   11.18453854],
                       [  13.80581557,   11.18453854,  209.89396417]])
                       
C_test = Tau_test.I

apatil

from __future__ import division
import numpy as np

# A boat is a dict with keys
# mass, center_of_mass, cm_depth, front_depth, back_depth, pitch, length, width
# units are radians (angles), meters (depths and lengths), tons (mass)
# Boats are idealized as rectangular prisms

water_density = 1

apatil

#!/usr/bin/env python

import pymc
from pymc import gp
from pymc.gp.cov_funs import matern,gaussian
from pylab import *

# Load some data generated from a GP with mean=0, scale=1, amp=1
xdata,ydata = loadtxt('train.txt', unpack=1)

apatil

#!/usr/bin/env python
import numpy as npy 
from pymc.gp import Mean, Covariance, Realization, observe, plot_envelope, NearlyFullRankCovariance, FullRankCovariance
from pymc.gp.cov_funs import matern #, thinplate1d
import matplotlib
matplotlib.rcParams['axes.facecolor']=[1,1,1]

__all__ = ['surface_mean', 'M', 'C']
def surface_mean(x, val):
    """docstring for parabolic_fun"""

apatil

import pymc as pm
import numpy as np

# FIXME: Need to store duplicates too, when jumps are rejected. That means some mechanism
# for making sure the history is full-rank needs to be employed.

class HistoryCovarianceStepper(pm.StepMethod):
    
    _state = ['n_points','history','tally','verbose']
    

apatil

from copy import copy
import theano as th
import theano.tensor as T

def isshared(node):
    "Is there a better way to do this?"
    return hasattr(node, 'update')
    
def unpack_nodes(expr):
    """

apatil

from scipy.sparse import linalg
import numpy as np
import map_utils

def dim2w(dim):
    return np.arange(-np.ceil((dim-1)/2.), np.floor((dim-1)/2.)+1)

class fractional_modified_laplacian(linalg.LinearOperator):
    
    def __init__(self, dims, kappa, alpha):

apatil

import numpy as np
from theano import tensor as T
import theano

x = T.dvector('x')
z = T.dscalar('z')
y = T.sum(T.cos(x*z))

x_test = np.random.normal(size=10)

apatil

#!/bin/bash

#
# template script for generating container for LXC 
# with username and password 'sense'
# and an ssh key for the root user.
#
# This script consolidates and extends the existing lxc ubuntu scripts
#