This is another post in our series What Machine-Learning Students Think/Like/Know/Are …, which we began during teaching our Machine-Learning module at Trinity College Dublin, Ireland.
Today, we present the machine-learning students’ experience and preferences of machine-learning libraries such as scikit-learn, Tensorflow, and Weka. This survey was conducted in one of the last lectures of our machine-learning module.
According to our survey, 72 of 73 students have used scikit-learn (98.6%). This is no surprise because we used scikit-learn as reference machine-learning library in our lectures and for the coursework. This means every student had to use scikit-learn. Or, more precisely, since all work was done in groups, every group had to use scikit-learn. The choice of using additional frameworks was up to the students. 31 students (42%) have worked with Tensorflow, 11 students (15%) with Weka, 8 students (11%) with Keras, 7 students (10%) with MLib for Spark, and 6 students (8%) with PyTorch. Other machine-learning libraries such as Caffe (2), Apache Mahout, Theano, KNIME, mlpack, Torch, MXNet, and deeplearn.js were used by very few or even no students.