The Signals Cookbook

This document will contain some solutions to common problems in Signals, and some clever bits of code that you can adapt for your own experiments.

Trial states
Etc.

Trial states

Here's how you work on data collected over a trial and reset this history on new trials.

% For this demonstration we create two signals, 'x' (e.g. an input device),
% and a 'newTrial' event.
[x, newTrial] = sig.test.create;

% Each new trial update the seed with an empty array, thus reinitializing
% our accumulated array.
seed = newTrial.then([]);
trialSamps = x.scan(@horzcat, seed);

A second slightly more memory controlled way of doing this is by using a buffer signal. Behind the scenes this initializes an array of a given size (in this example 1000 elements). We then simply create a signal to keep track of the current buffer index and slice the array at a different point each trial. You can pick any suffciently large number to initialze the buffer with. It should be larger than the number of samples you expect to collect per trial.

n = 1000; % Number of spaces in the buffer
hist = x.bufferUpTo(n); % Collect values of x into buffer
j = mod(x.map(1).scan(@plus,0), n); % Current index in buffer
i = j.at(newTrial); % Index at new trial
slice = iff(i < j, i:j, [i:n 1:j]); % deal with wrap-arounds
trialSamps = hist(slice);

Note that it's usually possible to avoid having to do this accumuate a heuristic or summary statistic that can be rest each trial. For instance if 'x' in the above example was a rotary encoder and you need the total displacement per trial, you could do this without storing the individual values in a buffer:

% Trial displacement is the difference between current position and
% position at trial start
displacement = x - x.at(newTrial);

% Trial distance is the sum of absolute position changes, resetting to 0
% at trial start
distance = x.delta().abs().scan(@plus, newTrial.then(0));

[start, choice, amt] = sig.test.create;
nPots = 3;
pots = cell(1,3);
for n = 1:nPots
  pots{n} = start.map(@(~) randi(1000));
end

f = @minus;

chosenPot = choice.selectFrom(pots{:}); % when i == 1, y = A, etc.
choiceHistory = choice.bufferUpTo(1000);

sum(choiceHistory == choice) * amt;

Etc.

Author: Miles Wells

v0.0.1

The Signals Cookbook

Contents

Trial states

Etc.