Saturday, March 8, 2014

Inputs are outputs and meta neural networks

Excited and out of breath, the employee meets the professor and says:

I’m happy for this opportunity to meet you.

The system uses non-trivial AI techniques neural networks and an expert system. The expert system is just a database but what’s interesting is the neural network whose activation functions are calibrated using a neural network itself. This was important because we reused the different segments of a neural network so the transition functions at each level of the hidden layers was important because they were both outputs to other NN or algorithms and inputs to the succeeding hidden layers. There was  a lot of activation functions to be designed and calibrated so we decided to create  a neural network to help us create those functions for the networks we are actually aiming to create.

And that’s the thing that I think warrants me a Master’s Degree.

No comments:

Post a Comment