Abstract: Feature extraction is a method that reduces the amount of information to efficiently describe data. It plays a very important role in classification and recognition. Feature extraction has, as input, a set of measured data and builds derived values (features) intended to be informative and non-redundant, facilitating the subsequent learning and generalization steps, and often leading to better human interpretations. Having irrelevant features in your data can often decrease the accuracy of predictive models, especially algorithms such as linear and logistic regression. Some advantages of feature extraction are: reducing overfitting, improving accuracy, and reducing training time. The examples will be developed in Python using some known libraries (e.g., scikit-learn, opencv).
During this workshop we will:
Compare different feature extraction methods (e.g., contour profiles, univariate selection, feature agglomeration, and principal component analysis).
Apply feature extraction to different types of data (e.g., images, text, sound).
Illustrate how the extracted features could be integrated within a machine learning model.
Prerequisites: All participants are expected to bring a laptop with a Mac, Linux, or Windows operating system that they have administrative privileges on. Familiarity with Python and machine learning techniques is encouraged. An RCC account is helpful, but not required.