While not the first of its kind, the Self-Organizing Map is a form of artificial neural network that became useful because of its use in visualizing and analyzing high-dimensional data on a low-dimensional scale. It gained its common name, the Kohonen Network, after a Finnish professor by the name of Teuvo Kohonen introduced it to the world .
What made it different from early neural network models is the fact that in these early neural network models, the self-organizing power was weak; thus this made the Kohonen Network crucial is that is composed no less than two interacting subsystems which are different in nature. Additionally, the whole point of learning in the SOM is to make different parts of the network react in a similar manner to a specific input pattern.
Since its development, the SOM has been cited in many different works making professor Kohonen the most cited Finnish professor in the world wherein the map is usually expressed in an abstract form. While this may be how it is usually used, the SOM can also be expressed in a purely mathematical form without the need to position it with underling components, neural or otherwise.
The article read also gave a few recommendations when applying the SOM to ensure that the work done is not only correct, but precise and understandable. These recommendations include but are not limited to:
- Checking the scale of the vector components
- Checking on the form of the array
- Being able to learn with a limited amount of training samples
- Using different qualities of learning samples
The article also goes on to explain that since its development, there are many different variations of the SOM. The reason behind this is the fact the SOM is adaptive in nature and even its basic principle can be adapted to be used to compute for and analyze subspace relations.
The second article read was about creating a cerebellum model that can run in real-time. The whole point of this (or what seemed to be the point of the research paper) is the current uses of having a cerebellum running in real time, the possible innovations it has both on humanity and robotics and pure scientific curiosity which is commonly the bedrock of more practical applications of science.
The paper goes on to explain the importance of a cerebellum, the origin of the idea and how they applied their experiments, the results of the experiment, the math used, the difference between the model they used and other models and lastly the potential future uses of the real-time cerebellum .
Neural Networks in the Field of Medicine
As the study of neural networks is part of studying the brain, it is not unexpected that studies, use and development of neural networks are closely related to the field of medicine. Since the first of its kind was developed, artificial neural networks (ANN) are used for a great number of things in the field of medicine and the field still has a great number of untapped potential that can, and are, being looked into. Depending on the ANN being used, the application varies; but what they do have in common is that they can be used for the following:
- As a diagnostic system
- ANN’s are used extensively in this field. They are currently useful as diagnostic system in detecting cancer and heart problems wherein its best asset is that when ANN’s are used in diagnostics, they are not affected by a number of different factors that can throw off the diagnosis such as the patient’s emotional state .
- In the field of biochemical analysis
- ANN’s are also currently being used to analyze blood and urine samples, check on the glucose level of a patient who is diabetic and even spot pathological disease such as tuberculosis .
- Image Analysis
- Medical imaging also uses artificial neural networks. In this field, ANN’s are used to identify chest x-rays. Classify tissue and vessels in an MRI (magnetic resonance images) and find out the skeletal age of an x-ray image .
- Development of Medicine
- ANN’s can also be used in developing new ways to treat different ailments and create new drugs for them. An example of this is creating new medicine for curing AIDS and treating cancer. It is also used in modeling bimolecular molecules .
Additionally, artificial neural networks have also been used in the field of medical statistics. An example of this is found in the work of the University of Oxford. ANN methods were used in determining the statistics of the survival rate of two different patients with cancer. The paper took into account the regressive problems of cancer, classification of a prognosis problem, analyzing survival rates of other patients with the same form of cancer, and how the neural networks fit into the scheme of things .
There are other ways to apply ANN’s in medicine like in the field of predictive medicine, but the real reason why ANN’s are used extensively in medicine is not because it was developed for it (as this was merely a side-effect of the ANN’s usefulness) but rather because of the key function of ANNs: processing huge and complicated amounts of date which doctors then use to make the best possible within the given situation for the patient in need ..
Another example of the use of ANN in medicine is its use in two simultaneous situations. Earlier, it was stated that ANN’s can be used to diagnose diseases; in 2010, an article was published where it showed that the application of ANN’s can be used to diagnose cancer using demographic data. The article was written with the idea that ANN’s can and have been used in accurate pre and post-clinical diagnosis and that the development of ANN’s is not just an accomplishment in medical science alone, but also in computer science .
Currently, the use of artificial neural networks has made the lives of many a great number of people easier. This is due in no part to the fact that ANN’s basically uncomplicated and simplifies what would usually be a great amount of data into something that can be read, understood and analyzed which in turn has led to more lives being saved in the field of medicine.
While there are other applications of ANN such as computer science, statistics and robotics, it still has its roots firmly in medical sciences as this is the field where it can be used to its maximum potential. Additionally, the use of ANN in other fields, such as robotics, still has the development of medical science as one of its goals.
Albu, A., & Ungureanu, L. Artificial Neural Network in Medicine. Politehnica University Timisoara.
Amato, F., Lopez, A., Peña-Méndez, E. M., Vaňhara, P., Hampl, A., & Havel, J. (2013). Artificial neural networks in medical diagnosis. Journal of .
Ganesan, N. D., Venkatesh, K. D., & Rama, M. A. (2010). Application of Neural Networks in Diagnosing Cancer Disease Using Demographic Data. International Journal of Computer Applications .
Grossi, E. (2011). Artificial Neural Networks and Predictive . Retrieved February 10, 2014, from Intechopen: http://www.intechopen.com/download/get/type/pdfs/id/14887
Kohonen, T. (1982). Self-Organized Formation of Topologically Correct Feature Maps. Biological Cybernetics , 59-69.
Kohonen, T., & Honkela, T. (n.d.). Kohonen network. Retrieved February 10, 2014, from Scholarpedia: http://www.scholarpedia.org/article/Kohonen_network
Ripley, B., & Ripley, R. (n.d.). Neural Networks as Statistical Methods in Survival. Artiﬁcial Neural Networks: Prospects for Medicine .
Siganos, D. (1995). Neural Networks in Medicine. Retrieved February 10, 2014, from http://www.doc.ic.ac.uk/~nd/surprise_96/journal/vol2/ds12/article2.html
Yamazaki, T., & Igarashi, J. (2013). Realtime cerebellum: A large-scale spiking network model of the cerebellum that runs in realtime using a graphics processing unit. Neural Networks , 103-111.