This paper presents a collection of in-class demonstrations for an introductory machine learning (ML) class. Each demonstration engages students actively in visualizing the behavior of a machine learning algorithm in order to build an intuitive understanding. These demonstrations are in direct contrast to purely slide- or whiteboard- based presentations of the same concepts by being student-paced and highly interactive, leveraging the physical space of the classroom. We developed demonstrations for six common ML methods: decision trees, k-nearest neighbors, the Perceptron, stochastic gradient descent (SGD), neural networks, and multi-armed bandits. Survey data from two semesters show that our demonstrations enhance student retention of and engagement with the material, relative to lectures without similar in-class demonstrations. Our demonstrations use readily available materials and student volunteers, making them easily reproducible for any educator seeking to complement their existing ML course.