What is this course about?
Since the dawn of the Artificial Neural Networks (ANNs), scholars and researchers have come up with various ways to facilitate the training process of the ANNs, for their desired applications. For example, if they needed an ANN to predict the number of sold cars in BMW Co., or the stock market price for the next month (Both Regression Problems), they came up with suitable error functions for regression (e.g., Sum Squared Error (SSE) function). It is important for machine learning researchers to LEARN the ways of developing an Error Function, which is tightly tied to the application at hand. Why? Because we frequently need to address novel applications in our projects/research and knowing the exact derivations and mathematical steps behind the scene is of utmost importance. Here is the list of what you will learn after you have finished this course:
- You will have a solid understanding of the famous Maximum Likelihood Estimation (MLE)
- You will know how one could use MLE, combine it with supervised learning, and produce suitable error functions for a regression/classification problem: namely the birth of the Sum Squared Error (SSE) and Cross-Entropy error function.
- MOST IMPORTANTLY, given a novel requirement by your application, you will be able to deeply analyze the problem at hand and develop your own error function, the minimization of which will result in accomplishing your novel application properly!
Enrol and Enjoy!