How it differs from SEM and TEM. Please provide any comparative study and papers regarding this.
Thanks in Advance!
Wish you all a Blessed Festive Season !
Kind Regards,
ANILA AJAYAN
Each specie has unique and peculiar characteristics and or properties. Therefore by training of the Artificial Neural network and generalization is achieved using different species; the ANN will identify any of such specie. However, it is important to note that ANN can not identify a specie it has not been trained with.
ANN, and many other classifiers, learns how to classify specimens through a "training set", which is a set of examples where an expert previously labeled the correct species.
This paper explains how some classifiers learn using a training set for bee species discrimination:
http://www.sciencedirect.com/science/article/pii/S0168169915000873
https://www.researchgate.net/publication/229087215_Specialized_Measures_of_Leaf_Outlines_and_Hierarchical_Perceptrons_in_an_Automatic_Identification_of_Plants
Article Specialized Measures of Leaf Outlines and Hierarchical Perce...
Thank you for the answers. So ANN can be useful if that particular organism's details are feed on it. Well, I would like to get more details on papers related to ANN in phtoplankton identification. Can anybody explain briefly.
Thank you
ANILA AJAYAN
Dear Anila Ajayan,
Thanks for your interesting question.
I have never developed any ANN tool to this application. But, If I might say and you know, there are many types of ANNs, and every type has a performance and works different.
Nevertheless, one of the main characteristic to classify ANNs consists of having the target value. That is to say, supervised or unsupervised ANN, respectively.
In the supervised case, a mathematical model is established between inputs and known outputs. This model depend on many important factors such as topology, transfer function, training function, type of data, and so on.
In the case of unsupervised ANN, the input values are classified by themselves looking for the minimum distance between them and other previously parameters fixed and that have to be optimized.
You can try a multilayer perceptron (supervised) and a SOM (unsupervised) ANNs. And according the results you can dig into this field.
Nevertheless, if you think that I can help you in looking for the best ANN that will solve your problem, it would be a pleasure for me.
Cheers,
José S. Torrecilla.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling