kcarnold

Compare colors by hue, saturation, or lightness. The sliders adjust the weights. It seems like the main effect is whether the weight is 0 or not.

The embedding is simply links to the 5 nearest neighbors, according to that metric.

kcarnold

This is a simple implementation of MDS. Dragging an item constrains it; double-clicking removes the constraint. The cost is shown in the top left.

The data is eurodist from R.

kcarnold

import numpy as np
import sys

def read_stored(num_dims, names_filename, data_filename):
    import pandas as pd
    names = pd.Index(line.strip() for line in open(names_filename))
    num_terms = len(names)
    data = np.memmap(data_filename, dtype=np.float32, mode='r', shape=(num_terms, num_dims))
    return names, data

kcarnold

.mturk-ari {
  border: 1px solid black;
  padding: 4px;
}

.mturk-ari th {
  padding-right: 5px;
  text-align: right;
}

kcarnold

def reconstruct_lkj_cholesky(ell, R2):
    num_items = len(R2) + 1
    L = np.zeros((num_items, num_items))
    L[0, 0] = 1.
    L[1, 0] = 2. * R2[0] - 1.0
    L[1, 1] = np.sqrt(1.0 - L[1, 0])

    start = 0
    for i in range(2, num_items):
        ell_row = ell[start:start+i]

kcarnold

import pandas as pd
import sys
filename, outfile = sys.argv[1:3]

responses = pd.read_csv(filename)

def format_row(row):
    return '\n'.join(["<p><b>{}</b><br>{}</p>".format(k, v.replace('\n', '<br>')) for k, v in row.iteritems()])

kcarnold

#
# SPI driver (using spidev device)
#
# Copyright (c) 2013  Kenneth Arnold <[email protected]>
# Copyright (c) 2007  MontaVista Software, Inc.
# Copyright (c) 2007  Anton Vorontsov <[email protected]>
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License.

kcarnold

#!/usr/bin/env python

import numpy as np
import sys

from six import iteritems
from six.moves import zip as izip
from six.moves import xrange
from itertools import chain, repeat, islice

kcarnold

from collections import OrderedDict
from boto.mturk.connection import MTurkConnection
from dateutil.parser import parse as dateparse

mtc = MTurkConnection(host='mechanicalturk.amazonaws.com')


def responses(hit_group_id):
    responses = []
    for hit in mtc.get_all_hits():

kcarnold

import numpy as np
from sklearn.covariance import graph_lasso
from sklearn.utils.extmath import pinvh

def compute_K(n, S, D):
    K = np.zeros((n,n))
    for a, b in S:
        K[a,b] = 1
        #K[b,a] = 1
    for a, b in D: