09.09.2014 Views

13th International Conference on Membrane Computing - MTA Sztaki

13th International Conference on Membrane Computing - MTA Sztaki

13th International Conference on Membrane Computing - MTA Sztaki

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

A case-study <strong>on</strong> the influence of noise to log-gain principles for flux dynamic<br />

discovery<br />

Using the log-gain theory requires that we know the initial values of all uxes<br />

and at least 1,000 steps of each time-series. [23] With data collected from the<br />

noisy time series we c<strong>on</strong>structed, we are easily able to supply this informati<strong>on</strong>.<br />

For our study, we used both the initial values of the uxes and 10,000 steps of<br />

our noisy time series data for the log-gain MetaPlab plugin.<br />

Our next step was to take the noisy time series data and import it into<br />

a model which c<strong>on</strong>tained no ux functi<strong>on</strong>s. The above model has had it's ux<br />

functi<strong>on</strong>s removed and the relevant data replaced with our noisy time series data.<br />

Such a model is what would typically be used with the log-gain theory, as the<br />

log-gain process should generate ux-dynamics by utilizing the time series data<br />

gathered from experimentati<strong>on</strong> and without making use of known ux functi<strong>on</strong>s.<br />

This model is a prime candidate for being used with the MetaPlab log-gain<br />

plugin. With noiseless data this process would be quickly veried; however, for<br />

our noisy data additi<strong>on</strong>al tests will be required. After importing all noisy data<br />

and the initial values for the ux functi<strong>on</strong>s, we were ready to begin the log-gain<br />

process.<br />

We made use of the following set of tuners in using our model with the loggain<br />

process. Tuner selecti<strong>on</strong> was a simple task as we knew the ux functi<strong>on</strong>s<br />

before hand (recall that we simply assumed some ux functi<strong>on</strong>s). As such, tuner<br />

selecti<strong>on</strong> was a simple process of determining which reacti<strong>on</strong>s were governed by<br />

which substances.<br />

R 0 = A, B, C<br />

R 1 = B,C<br />

R 2 = C, A<br />

R 3 = B<br />

R 4 = C<br />

As menti<strong>on</strong>ed above, we made use of the MetaPlab log-gain plugin to apply<br />

log-gain theory to our model. This plugin requires the initial values of all uxes<br />

and at least 1,000 steps of each time-series. The time series data used is the<br />

10,000 step noisy data we generated. The reacti<strong>on</strong> set was input into the plugin<br />

with a good Covering Oset Log Gain Property [10], and a polynomial was then<br />

generated through the plugin by selecting an appropriate substance. Descripti<strong>on</strong>s<br />

for the polynomials were given at the time of generati<strong>on</strong>. The reacti<strong>on</strong>s were set<br />

up in the following manner: R0 covered substance A, R3 covered substance B,<br />

and R2covered substance C. After all tuners and coverings were input into<br />

MetaPlab, we ran the plugin. We also used a linear regressi<strong>on</strong> plugin to apply<br />

curve tting to the clean noise eect and retrieve the ux functi<strong>on</strong>s. This <strong>on</strong>ly<br />

required that we make use of the ux functi<strong>on</strong>s and the initial values of the<br />

various time series. The time series can either be imported from external les by<br />

the log-gain plugin or computed by using the dynamic computati<strong>on</strong> plugin.<br />

The log-gain plugin applied log-gain theory to the noisy time series data, but<br />

to better understand our results, we took <strong>on</strong>e last step. We noticed that our<br />

75

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!