Exercise Sheet 4 – Solutions Topics : Artificial Neural Networks : Use Cases

Implementations of neural networks In this week exercise, I’ve selected four specific articles using artificial neural networks: 1. Artificial neural network-based segmentation and apple grading by machine vision, Devrim Unay and Bernard Gosselin. 2. Neural Network-Based Face Detection, Henry A. Rowley, Shumeet Baluja, and Takeo Kanade. 3. A Transportable Neural-Network Approach to Autonomous Vehicle Following, Nasset Kehtarnavaz, Norman Grisworld, Kelly Miller, Paul Lescoe. 4. On the Trajectory Tracking Control of Industrial SCARA Robot Manipulators, Antonio Visioli, and Giovanni Legnani. For each of these articles, the exercise is to identify the main characteristics of the neural network used: (a) What is the kind of network used? Is it a classifier or an approximator? (b) What are the input and output of the neural network? (c) How is the neural network trained? (d) How is the performance of the network evaluated? (e) What motivated the choice of a neural network implementation?