PACE

PACE: Parameterization & Analysis of Conduit Edges William Farmer - 2015

What is it

This tool is a way to examine recorded lab data from conduit experiments and determine the usability based on the straightness of each conduit as well as the noise.

Arguments

usage: python3 linefit.py [-h] [-p] [-pd] [-a] [-mt] [-m MODEL] [-nnk NNK]
                          [-t TIME] [-d DATASTORE] [-pds] [-pdss]
                          [F [F ...]]

Parameterize and analyze usability of conduit edge data

positional arguments:
  F                     File(s) for processing. Each file has a specific
                        format: See README (or header) for specification.

optional arguments:
  -h, --help            show this help message and exit
  -p, --plot            Create Plot of file(s)? Note, unless --time flag used,
                        will plot middle time.
  -pd, --plotdata       Create plot of current datastore.
  -a, --analyze         Analyze the file and determine Curvature/Noise
                        parameters. If --time not specified, will examine
                        entire file. This will add results to datastore with
                        false flags in accept field if not provided.
  -mt, --machinetest    Determine if the times from the file are usable based
                        on supervised learning model. If --time not specified,
                        will examine entire file.
  -m MODEL, --model MODEL
                        Learning Model to use. Options are ["nn", "svm",
                        "forest", "sgd"]
  -nnk NNK, --nnk NNK   k-Parameter for k nearest neighbors. Google it.
  -t TIME, --time TIME  Time (column) of data to use for analysis OR plotting.
                        Zero-Indexed
  -d DATASTORE, --datastore DATASTORE
                        Datastore filename override. Don't do this unless you
                        know what you're doing
  -pds, --printdata     Print data
  -pdss, --printdatashort
                        Print data short

Datastore Format

learning_data = [{filehash:[[trial_index, curvature, noise,
                            range, domain, accept, viscosity]
                    ,...],...},...]

The filehash is a SHA512 hash of the file contents read in 4kb chunks to prevent hash collision.

New values can be added to this file manually, just take care that SHA512 is used for these new key-value pairs.

Incoming Data Format

.mat file, with the following format:

{left_edges:2d_array,
 right_edges:2d_array,
 times:1d_array,
 accept:1d_array,
 viscosity:1d_array,
 ratio:float}

left and right edge arrays are 2 dimensional arrays where each column corresponds to the time that that "photo" was taken. These are float values corresponding to pixel heights.

The time array corresponds to the exact times that each column photo was taken. These are float time values.

accept is an array that corresponds to whether or not each column was accepted. These are integers 0 or 1. Any other values will be assumed to be 0.

viscosity is an array of viscosity values. Assumed to be 1 if not provided.

ratio is a float value that corresponds to the pixel-per-inch ratio.