In this course students are introduced to the architecture of deep neural networks, algorithms that are developed to extract high-level feature representations of data. In addition to theoretical foundations of neural networks, including backpropagation and stochastic gradient descent, students get hands-on experience building deep neural network models with Python. Topics covered in the course include image classification, time series forecasting, text vectorization (tf-idf and word2vec), natural language translation, speech recognition, and deep reinforcement learning. Students learn how to use application program interfaces (APIs), such as TensorFlow and Keras, for building a variety of deep neural networks: convolutional neural network (CNN), recurrent neural network (RNN), self-organizing maps (SOM), generative adversarial network (GANs), and long short-term memory (LSTM). Some of the models will require the use of graphics processing unit (GPU) enabled Amazon Machine Images (AMI) in Amazon Web Services (AWS) Cloud.
Harvard Division of Continuing Education