eric-tramel

import scipy as sci

## Check Python Version
import sys

assert sys.version.find('3.6') > -1

## Check numpy
import numpy as np
from math import sqrt

eric-tramel

/*
    In this program we will review the queue data structre from the perspective 
    of C.

    However, technically, the implementation here is for a *circular array* 
    which we can access from either end, as well as popping from either end.
    The main feature of this implementation is the fact that the framework is
    general to the type of variable to be stored within the circular array.

    Obviously, one might attempt an implementation using linked lists, which

eric-tramel

function wmat = wcell2mat(wcell)
    L = length(wcell);
    
    bb = wcell{end};
    [rlo,clo] = size(bb);
    
    wmat = bb;
    
    for i=(L-1):-1:1
        wmat = [wmat, wcell{i}{1};

eric-tramel

clear;

%% File Paths
fname = 'AllCIFAR.h5';
addpath(genpath('/Google Drive/Research/code/matlab/toolboxes/WaveletSoftware'));

%% Read in the HDF5 information
hi = hdf5info(fname);
X = hdf5read(hi.GroupHierarchy(1).Datasets(1));

eric-tramel

function X = generate_synthetic_tomo_data(N,p,seed)
% GENERATE_SYNTHETIC_TOMO_DATA Create an NxN circle-masked synthetic image based
% on E.G.'s BP for Tomography package: https://github.com/eddam/bp-for-tomo. 

% def generate_synthetic_data(l_x=128, seed=None, crop=True, n_pts=25):
 %    """
 %    Generate synthetic binary data looking like phase separation

 %    Parameters
 %    ----------

eric-tramel

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Testing CuBLAS and CUDArt for Julia\n",
    "After finally getting NVCC to work on OSX, we can start using the CUDA-themed BLAS packages written for Julia. In this notebook we will document how to utilize the necessary datatypes and show comparisons between the CPU and GPU implementations of common BLAS functions."
   ]
  },

eric-tramel

2015-09-21 13:23:35 +0200

make
-C
contrib
-f
repackage_system_suitesparse4.make
prefix=/usr/local/Cellar/julia/HEAD
USE_BLAS64=0
FC=/usr/local/bin/gfortran

eric-tramel

function x = eric_positive_armijo(x0,f,g,h)
% Calculate a Newton-Step type minimization using the given 
% f function, its gradient g, and its hessian (second order)
% h.

scalar_mode = false;
if isscalar(x0)
	scalar_mode = true;
end

eric-tramel

classdef Prior
    % This class contains all the prior-dependent functions including learnings
    
    properties
        av_mess; av_mess_old; var_mess; var_mess_old; R; S2; rho; learn; N; alpha; func; dump_learn; t; method; param_1; param_2; param_3; param_4;
        % Gaussian sparse prior : p(x) ~ (1 - rho) * delta(x) + rho / sqrt(2 * pi * var_gauss) * exp(-(x - m_gauss)^2 / (2 * var_gauss) ) : param_1 = m_gauss; param_2 = var_gauss;
        % Gaussian sparse prior enforcing value inside a symetric interval : p(x) ~ [(1 - rho) * delta(x) + rho / sqrt(2 * pi * var_gauss) * exp(-(x - m_gauss)^2 / (2 * var_gauss) )] * I(|x| < cut) : param_1 = m_gauss; param_2 = var_gauss; param_3 = cut;
        % Positive Gaussian sparse prior : p(x) ~ (1 - rho) * delta(x) + rho / sqrt(2 * pi * var_gauss) * exp(-(x - m_gauss)^2 / (2 * var) ) * I(x > 0) : param_1 = m_gauss; param_2 = var_gauss;
        % Mixture of two gaussians : p(x) ~ (1 - rho) * exp(-(x - m_1)^2 / (2 * var_1) ) / sqrt(2 * pi * var_1) + rho * exp(-(x - m_

eric-tramel

% Specify image
im_size = [128,128];
X = double(imread('peppers.png'));
X = imresize(X,im_size);

% Grab some kind of filter/PSF
PSF_dim = [4,4];
PSF = ones(PSF_dim);
PSF = PSF ./ norm(PSF(:));