Prior to this research project, my only experience coding included two introductory courses. The first was Introduction to Robotics, where I programmed Lego RXT and NXT with a visual programming language called Robo Lab. The second was Software Design and Implementation where I learned the Fundamentals of Programming using Python.

   So to say I felt under prepared is an understatement. When Dr.Crandall and his doctoral student told me to program a multi-layer perception from scratch, my initial reaction was a question. "Just open up a blank script and program," I asked in disbelief. They told me yes and said that I could only refer to research articles for help, no previously written code.

   It took me about a week a reading research papers to understand enough of the concept of neural networks to start programming. The neural network itself, was fairly straight forward to program, but the learning algorithm was a different story. Still, I finished the program all by myself. It was slow and inefficient because my understanding of stochastic gradient descent was a bit shaky, but it worked.

   After that, the doctoral student helped me optimize the learning algorithm and I trained one neural network to categorize three different types of Iris flowers and another to recognize hand written digits from the MNIST data set.

   After writing a few programs to train the network and waiting about a week for it to train, I was excited to see the network perform with high accuracy. 

   This initial experience with Machine Learning and Artificial Intelligence sparked my interest in the subject. Now, its a major part of my academic aspirations in robotics.