I have been doing lot of work in the field of artificial intelligence over last 25 years since 1993. I thought It would be good to write few summary post about my journey and experiences.

I had interests in Hypnotism, telepathy etc since 1983 when I read a book on that subject. I kept wondering how telepathy could work. But over time I forgot to pursue. And then in the winter of 1990, during my first year of M.Tech (Applied Geophysics) in IIT Roorkee, I happened to have discussion on Telepathy with Dushyant Kumar, a friend and batchmate in IIT Roorkee who was doing MSc in Biotechnology. In that discussion he mentioned about neurons, neurotransmitters, and various kinds of waves (alpha, beta, and gamma) that are there in the brain. I made quick but faulty conclusion that these alpha, beta. and gamma waves are EM waves and so have possibility to enable Telepathic communication. Later during my literature survey I came to know that these are actually electrical waves.

I did not do much on the subject for nearly two years but curiosity remained and in free time I kept studying various books and journals in library. In Biotech journals I read good amount of stuff on neurons and neural networks. Back then there was no Google to search and find out if there is possibility of any artificial neural network. And then one day I saw an article Location of subsurface target in geophysical data using neural networks published in Geophysics on the application of artificial neural network in the field of Geophysics. I read it many times and also found and read other referenced articles. It is then I realized that my friends in computer science branch can be of help to discuss and understand more about it. Fortunately, I had a great friend, Now I remember only his surname - Nigam, with whom I discussed for guidance. And he actually had worked on it and gave me the photocopied version of Parallel distributed processing: explorations in the microstructure of cognition. Volume 1. Foundations by D.E. Rumelhart. After reading few chappers of this book I decided that I will do my M.Tech Dissertation in the field of Application of Artificial Neural Network for Geophysics. After discussing with my department guide Professor Sri Niwas, Shanti Swarup Bhatnagar Award winner, I searched for a co-guide from computer science department. And I was lucky to have Professor Anurag Tyagi (Phylum) who was pioneer researcher in the field of artificial neural networks at that time.

The topic of my dissertation was "Artificial Neural Network Application for Geophysical Interpretation Using Back-propagation Algorithm". I received highest marks and gold medal for doing this research work. It was later published in Journal of Geophysics sometime in 1995.

There are many more interesting stories while doing this work but I am skipping them to keep it brief and relevant for scientific read only. Still following events are worth mentioning:

1. I used ForTran to code the methodology. In my code I was feeding and updating weights of the neural network sample by sample. This was not helping network to learn and reduce error. I kept doing it for nearly a month then one day I decided to feed all the training pairs first and then do the weight update. I was surprised and happy to see the network error was reducing. After that I took only a month to wind up my work.
2. During dissertation I was asked two questions by Prof P. K. Gupta. That what is the source of nonlinearity in neural network and I answered that Sigmoidal functions used for squashing the output in neurons enable non-linearity, which i realized only when the question was asked. And second question was why I made my own Finite Difference Forward modelling program when they are available. And my humble answer was that I tried them but could not make them work so decided to write my own code.

In 1993, I qualified CSIR Exam to carry out government sponsored research. I took up the topic of Application of Artificial Neural Network to do inversion of Electromagnetic data to predict the depth, size, and conductivity of subsurface anomaly. For some personal reasons I left the research work and joined Industry in 1995.

Keep Steeming.

Relevant comments will be upvoted