hello by

either supplementing from the university medical college crackle bolt

a little attention to the paper entitled in we should be the relationship to generalize

model based on artificial on it

if you should be able correlation i d c is classical approach it's a mathematical

model

designed to predict the in-vehicle measured from energy profile based on the in vitro measure

dissolution provide of that particular trial of interest

this formula is used by pharmaceutical industry for the costs at purposes

because once it is established it allows to reduce the demand of prospective buyers

this is the classical approach of i'm a satellite based on instructions of it

you have or is it linear problem were drug fraction of sort of be said

and via and independent variable dry fraction within b

for the last are you have the results ready for computations measured in the lower

bound for czech

for the for you don't have text is its measurement of this profile therefore you

need to create based on numerical approaches one of them the most commonly used it's

the deconvolution

and the deconvolution methods unfortunately in order to be precise and accurate demands an additional

in table five which is usually derived from the

intravenous restriction particular drug

this situation

it's not very comfortable for pharmaceutical industry goes

first it requires an additional by a slight and second us for certain cases

okay

i know administration is sometimes impostor

for particular so we would like to propose some solution distribution based on artificial neural

networks

computational intelligence tools

eight to create very complicated relationships

and therefore to provide direct relationship idea of our model without a deconvolution phase and

these obstacles

so that we shall do not show how we do

we put two into the input of the neural network to dissolution provide accompanied by

the chemical formula of the active pharmaceutical entreaty and icsi and encoded in so called

monocular descriptors

and also we introduce some parameters of the dissolution test

to the output in order to train the one that's work we put a presented

a corresponding from canada provides derived from each range

this is an example all the prediction power system an average prediction

we can see it's not it's not perfectly precise

it's still able to need a job of course of the of the kinetic profile

and it is also a able to predict the range of considerations

which is very important feature the small

for more details i would like to invite you to drive design development and their

actual

i we should have really thank you for that image