We know Deep neural network is quite famous for it's ability to work in the same way as human brain does. It can detect any kind of object or patterns within an image or RAW data. But what is the actual step by step process and what it actually execute within it to finalize detection process. Need a brief description as well as suitable references will be better for clearing my concept.
You might want to look at the paper in the link. It describes deep learning and convolutional networks.
Conference Paper Multimedia data mining using deep learning
Dear Sir Dr. Peter Wlodarczak,
Thank you for your reply. I have gone through the work you have mentioned. it's really a standard work but i need some works where deep neural network is depicted using lots of mathematical functions as well as equations. moreover, i need to know mathematical derivation of DNN. Thus, it will be helpful for me if you could mention few more works on DNN.
You can start with the works of:
I would need more information based on whether you choose 1 or 2 to give you more specific references on the subject
Dear Sir Dr. Arturo Geigel,
Thank you so much for your reply. Since i'm working with CNN, it will be better if you kindly mention some papers for option 2.
You could start with Yoshua Bengio's work:
Learning deep architectures for AI
http://www.iro.umontreal.ca/~bengioy/papers/ftml_book.pdf
look at page 43 that will give you an introduction and references to convolutional neural networks and their recent migration to RBM;s and DBM's. Though I would suggest that before diving into those look at the earlier references on CNN's like LeCun's and Bengios:
Gradient-Based Learning Applied to Document Recognition
This is one of the earlier works on CNN's and you will get insight on the earlier design decisions before moving on to more abstract papers. Obtaining the mathematical formulations for your specific work will depend on the architecture that you are contemplating. Once you read the papers referenced in the book and Lecun;s paper and have your design, we can discuss how to dig deeper for mathematical formulations based on specifics (for example, LeNet 5 contains in the final layer an RBF which by itself has been proven to mathematically be a universal approximator and has its own subset of references).
https://www.slideshare.net/mohamedloey/deep-learning-overview-of-my-work-ii
Deep learners use back-propagation for capital assignment and gradient descent for training. There is no evidence that any such mechanism in the brain exists. Deep learners are at best inspired by nature.
- Badges
- Science topic