Intendierte Lernergebnisse
Neural networks, deep learning (DL), and computational neuroscience have different applications in text categorization, e.g., spam filtering, fraud detection, optical character recognition, machine vision, e.g., face detection, licenses plate recognition, advanced driver assistance systems, natural-language processing, e.g., spoken language understanding, market segmentation, e.g., predict if a customer will get a credit, and bioinformatics, e.g., classify proteins or lipidomes according to their function. The material covered in this course expands upon what was covered in course 700.395, which is titled Data Mining and Neurocomputing. The following are examples of methods that will be covered: The principles of artificial neural networks (NN), Convolutional neural networks (CNN), Recurrent neural networks (RNN), Generative models, e.g., autoencoders, variational autoencoders, generative adversarial network (GAN).The theoretical foundations of the approaches that will be explored in the class will place a strong emphasis on application and modeling techniques using different datasets. The result of the learning is a solid understanding of deep learning fundamentals. The lecture will guide to transferring of the acquired knowledge to solve classification problems for industry and research, (c) the basics of computational neuroscience, and (d) show some use-cases and exciting applications from the state-of-the-art.
Lehrmethodik
Lecturesclassroom exercises, quiz and activities.EvaluationProject 50% + Quiz (20%) + Assignments (20%) + Class activity (10%)
Inhalt/e
Data preprocessing / data augmentationUnsupervised Learning and ClusteringDeep Learning (multilayer perceptron, convolutional models, recurrent models)Time series forecast (long-short-term-memory and gated recurrent units)Spiking neural networksDeep learning libraries (torch, theano, keras, tensorflow...etc.)Evaluation Metrics
Erwartete Vorkenntnisse
Desirable prerequisites:CalculusLinear algebraCoding skills in python
Literatur
Ian Goodfellow, Yoshua Bengio, Aaron Courville, Deep Learning, MIT Press, 2016.ISBN-13: 978-0262035613 ISBN-10: 0262035618