# Search result: Catalogue data in Spring Semester 2019

DAS in Data Science | ||||||

Specialisation Track | ||||||

Hardware for Machine Learning | ||||||

Number | Title | Type | ECTS | Hours | Lecturers | |
---|---|---|---|---|---|---|

227-0150-00L | Systems-on-chip for Data Analytics and Machine Learning Previously "Energy-Efficient Parallel Computing Systems for Data Analytics" | W | 6 credits | 4G | L. Benini | |

Abstract | Systems-on-chip architecture and related design issues with a focus on machine learning and data analytics applications. It will cover multi-cores, many-cores, vector engines, GP-GPUs, application-specific processors and heterogeneous compute accelerators. Special emphasis given to energy-efficiency issues and hardware-software techniques for power and energy minimization. | |||||

Objective | Give in-depth understanding of the links and dependencies between architectures and their energy-efficient implementation and to get a comprehensive exposure to state-of-the-art systems-on-chip platforms for machine learning and data analytics. Practical experience will also be gained through practical exercises and mini-projects (hardware and software) assigned on specific topics. | |||||

Content | The course will cover advanced system-on-chip architectures, with an in-depth view on design challenges related to advanced silicon technology and state-of-the-art system integration options (nanometer silicon technology, novel storage devices, three-dimensional integration, advanced system packaging). The emphasis will be on programmable parallel architectures with application focus on machine learning and data analytics. The main SoC architectural families will be covered: namely, multi and many- cores, GPUs, vector accelerators, application-specific processors, heterogeneous platforms. The course will cover the complex design choices required to achieve scalability and energy proportionality. The course will will also delve into system design, touching on hardware-software tradeoffs and full-system analysis and optimization taking into account non-functional constraints and quality metrics, such as power consumption, thermal dissipation, reliability and variability. The application focus will be on machine learning both in the cloud and at the edges (near-sensor analytics). | |||||

Lecture notes | Slides will be provided to accompany lectures. Pointers to scientific literature will be given. Exercise scripts and tutorials will be provided. | |||||

Literature | John L. Hennessy, David A. Patterson, Computer Architecture: A Quantitative Approach (The Morgan Kaufmann Series in Computer Architecture and Design) 6th Edition, 2017. | |||||

Prerequisites / Notice | Knowledge of digital design at the level of "Design of Digital Circuits SS12" is required. Knowledge of basic VLSI design at the level of "VLSI I: Architectures of VLSI Circuits" is required | |||||

227-0155-00L | Machine Learning on Microcontrollers Registration in this class requires the permission of the instructors. Class size will be limited to 25. Preference is given to students in the MSc EEIT. | W | 3 credits | 2G | M. Magno, L. Benini | |

Abstract | Machine Learning (ML) and artificial intelligence are pervading the digital society. Today, even low power embedded systems are incorporating ML, becoming increasingly “smart”. This lecture gives an overview of ML methods and algorithms to process and extract useful near-sensor information in end-nodes of the “internet-of-things”, using low-power microcontrollers/ processors (ARM-Cortex-M; RISC-V) | |||||

Objective | Learn how to Process data from sensors and how to extract useful information with low power microprocessors using ML techniques. We will analyze data coming from real low-power sensors (accelerometers, microphones, ExG bio-signals, cameras…). The main objective is to study in details how Machine Learning algorithms can be adapted to the performance constraints and limited resources of low-power microcontrollers. | |||||

Content | The final goal of the course is a deep understanding of machine learning and its practical implementation on single- and multi-core microcontrollers, coupled with performance and energy efficiency analysis and optimization. The main topics of the course include: - Sensors and sensor data acquisition with low power embedded systems - Machine Learning: Overview of supervised and unsupervised learning and in particular supervised learning (Bayes Decision Theory, Decision Trees, Random Forests, kNN-Methods, Support Vector Machines, Convolutional Networks and Deep Learning) - Low-power embedded systems and their architecture. Low Power microcontrollers (ARM-Cortex M) and RISC-V-based Parallel Ultra Low Power (PULP) systems-on-chip. - Low power smart sensor system design: hardware-software tradeoffs, analysis, and optimization. Implementation and performance evaluation of ML in battery-operated embedded systems. The laboratory exercised will show how to address concrete design problems, like motion, gesture recognition, emotion detection, image and sound classification, using real sensors data and real MCU boards. Presentations from Ph.D. students and the visit to the Digital Circuits and Systems Group will introduce current research topics and international research projects. | |||||

Lecture notes | Script and exercise sheets. Books will be suggested during the course. | |||||

Prerequisites / Notice | Prerequisites: C language programming. Basics of Digital Signal Processing. Basics of processor and computer architecture. Some exposure to machine learning concepts is also desirable | |||||

Image Analysis & Computer Vision | ||||||

Number | Title | Type | ECTS | Hours | Lecturers | |

227-0391-00L | Medical Image AnalysisBasic knowledge of computer vision would be helpful. | W | 3 credits | 2G | E. Konukoglu, M. A. Reyes Aguirre, C. Tanner | |

Abstract | It is the objective of this lecture to introduce the basic concepts used in Medical Image Analysis. In particular the lecture focuses on shape representation schemes, segmentation techniques, machine learning based predictive models and various image registration methods commonly used in Medical Image Analysis applications. | |||||

Objective | This lecture aims to give an overview of the basic concepts of Medical Image Analysis and its application areas. | |||||

Prerequisites / Notice | Prerequisites: Basic concepts of mathematical analysis and linear algebra. Preferred: Basic knowledge of computer vision and machine learning would be helpful. The course will be held in English. | |||||

227-1034-00L | Computational Vision (University of Zurich) No enrolment to this course at ETH Zurich. Book the corresponding module directly at UZH. UZH Module Code: INI402 Mind the enrolment deadlines at UZH: https://www.uzh.ch/cmsssl/en/studies/application/mobilitaet.html | W | 6 credits | 2V + 1U | D. Kiper | |

Abstract | This course focuses on neural computations that underlie visual perception. We study how visual signals are processed in the retina, LGN and visual cortex. We study the morpholgy and functional architecture of cortical circuits responsible for pattern, motion, color, and three-dimensional vision. | |||||

Objective | This course considers the operation of circuits in the process of neural computations. The evolution of neural systems will be considered to demonstrate how neural structures and mechanisms are optimised for energy capture, transduction, transmission and representation of information. Canonical brain circuits will be described as models for the analysis of sensory information. The concept of receptive fields will be introduced and their role in coding spatial and temporal information will be considered. The constraints of the bandwidth of neural channels and the mechanisms of normalization by neural circuits will be discussed. The visual system will form the basis of case studies in the computation of form, depth, and motion. The role of multiple channels and collective computations for object recognition will be considered. Coordinate transformations of space and time by cortical and subcortical mechanisms will be analysed. The means by which sensory and motor systems are integrated to allow for adaptive behaviour will be considered. | |||||

Content | This course considers the operation of circuits in the process of neural computations. The evolution of neural systems will be considered to demonstrate how neural structures and mechanisms are optimised for energy capture, transduction, transmission and representation of information. Canonical brain circuits will be described as models for the analysis of sensory information. The concept of receptive fields will be introduced and their role in coding spatial and temporal information will be considered. The constraints of the bandwidth of neural channels and the mechanisms of normalization by neural circuits will be discussed. The visual system will form the basis of case studies in the computation of form, depth, and motion. The role of multiple channels and collective computations for object recognition will be considered. Coordinate transformations of space and time by cortical and subcortical mechanisms will be analysed. The means by which sensory and motor systems are integrated to allow for adaptive behaviour will be considered. | |||||

Literature | Books: (recommended references, not required) 1. An Introduction to Natural Computation, D. Ballard (Bradford Books, MIT Press) 1997. 2. The Handbook of Brain Theorie and Neural Networks, M. Arbib (editor), (MIT Press) 1995. | |||||

252-0579-00L | 3D Vision | W | 4 credits | 3G | M. Pollefeys, V. Larsson | |

Abstract | The course covers camera models and calibration, feature tracking and matching, camera motion estimation via simultaneous localization and mapping (SLAM) and visual odometry (VO), epipolar and mult-view geometry, structure-from-motion, (multi-view) stereo, augmented reality, and image-based (re-)localization. | |||||

Objective | After attending this course, students will: 1. understand the core concepts for recovering 3D shape of objects and scenes from images and video. 2. be able to implement basic systems for vision-based robotics and simple virtual/augmented reality applications. 3. have a good overview over the current state-of-the art in 3D vision. 4. be able to critically analyze and asses current research in this area. | |||||

Content | The goal of this course is to teach the core techniques required for robotic and augmented reality applications: How to determine the motion of a camera and how to estimate the absolute position and orientation of a camera in the real world. This course will introduce the basic concepts of 3D Vision in the form of short lectures, followed by student presentations discussing the current state-of-the-art. The main focus of this course are student projects on 3D Vision topics, with an emphasis on robotic vision and virtual and augmented reality applications. | |||||

263-3710-00L | Machine Perception Number of participants limited to 150. | W | 5 credits | 2V + 1U + 1A | O. Hilliges | |

Abstract | Recent developments in neural networks (aka “deep learning”) have drastically advanced the performance of machine perception systems in a variety of areas including computer vision, robotics, and intelligent UIs. This course is a deep dive into deep learning algorithms and architectures with applications to a variety of perceptual tasks. | |||||

Objective | Students will learn about fundamental aspects of modern deep learning approaches for perception. Students will learn to implement, train and debug their own neural networks and gain a detailed understanding of cutting-edge research in learning-based computer vision, robotics and HCI. The final project assignment will involve training a complex neural network architecture and applying it on a real-world dataset of human activity. The core competency acquired through this course is a solid foundation in deep-learning algorithms to process and interpret human input into computing systems. In particular, students should be able to develop systems that deal with the problem of recognizing people in images, detecting and describing body parts, inferring their spatial configuration, performing action/gesture recognition from still images or image sequences, also considering multi-modal data, among others. | |||||

Content | We will focus on teaching: how to set up the problem of machine perception, the learning algorithms, network architectures and advanced deep learning concepts in particular probabilistic deep learning models The course covers the following main areas: I) Foundations of deep-learning. II) Probabilistic deep-learning for generative modelling of data (latent variable models, generative adversarial networks and auto-regressive models). III) Deep learning in computer vision, human-computer interaction and robotics. Specific topics include: I) Deep learning basics: a) Neural Networks and training (i.e., backpropagation) b) Feedforward Networks c) Timeseries modelling (RNN, GRU, LSTM) d) Convolutional Neural Networks for classification II) Probabilistic Deep Learning: a) Latent variable models (VAEs) b) Generative adversarial networks (GANs) c) Autoregressive models (PixelCNN, PixelRNN, TCNs) III) Deep Learning techniques for machine perception: a) Fully Convolutional architectures for dense per-pixel tasks (i.e., instance segmentation) b) Pose estimation and other tasks involving human activity c) Deep reinforcement learning IV) Case studies from research in computer vision, HCI, robotics and signal processing | |||||

Literature | Deep Learning Book by Ian Goodfellow and Yoshua Bengio | |||||

Prerequisites / Notice | This is an advanced grad-level course that requires a background in machine learning. Students are expected to have a solid mathematical foundation, in particular in linear algebra, multivariate calculus, and probability. The course will focus on state-of-the-art research in deep-learning and will not repeat basics of machine learning Please take note of the following conditions: 1) The number of participants is limited to 150 students (MSc and PhDs). 2) Students must have taken the exam in Machine Learning (252-0535-00) or have acquired equivalent knowledge 3) All practical exercises will require basic knowledge of Python and will use libraries such as TensorFlow, scikit-learn and scikit-image. We will provide introductions to TensorFlow and other libraries that are needed but will not provide introductions to basic programming or Python. The following courses are strongly recommended as prerequisite: * "Visual Computing" or "Computer Vision" The course will be assessed by a final written examination in English. No course materials or electronic devices can be used during the examination. Note that the examination will be based on the contents of the lectures, the associated reading materials and the exercises. | |||||

Neural Information Processing | ||||||

Number | Title | Type | ECTS | Hours | Lecturers | |

227-0395-00L | Neural Systems | W | 6 credits | 2V + 1U + 1A | R. Hahnloser, M. F. Yanik, B. Grewe | |

Abstract | This course introduces principles of information processing in neural systems. It covers basic neuroscience for engineering students, experimental techniques used in studies of animal behavior and underlying neural mechanisms. Students learn about neural information processing and basic principles of natural intelligence and their impact on efforts to design artificially intelligent systems. | |||||

Objective | This course introduces - Methods for monitoring of animal behaviors in complex environments - Information-theoretic principles of behavior - Methods for performing neurophysiological recordings in intact nervous systems - Methods for manipulating the state and activity in selective neuron types - Methods for reconstructing the synaptic networks among neurons - Information decoding from neural populations, - Sensorimotor learning, and - Neurobiological principles for machine learning. | |||||

Content | From active membranes to propagation of action potentials. From synaptic physiology to synaptic learning rules. From receptive fields to neural population decoding. From fluorescence imaging to connectomics. Methods for reading and manipulation neural ensembles. From classical conditioning to reinforcement learning. From the visual system to deep convolutional networks. Brain architectures for learning and memory. From birdsong to computational linguistics. | |||||

Prerequisites / Notice | Before taking this course, students are encouraged to complete "Bioelectronics and Biosensors" (227-0393-10L). As part of the exercises for this class, students are expected to complete a (python) programming project to be defined at the beginning of the semester. | |||||

227-0973-00L | Translational Neuromodeling | W | 8 credits | 3V + 2U + 1A | K. Stephan | |

Abstract | This course provides a systematic introduction to Translational Neuromodeling (the development of mathematical models for diagnostics of brain diseases) and their application to concrete clinical questions (Computational Psychiatry/Psychosomatics). It focuses on a generative modeling strategy and teaches (hierarchical) Bayesian models of neuroimaging data and behaviour, incl. exercises. | |||||

Objective | To obtain an understanding of the goals, concepts and methods of Translational Neuromodeling and Computational Psychiatry/Psychosomatics, particularly with regard to Bayesian models of neuroimaging (fMRI, EEG) and behavioural data. | |||||

Content | This course provides a systematic introduction to Translational Neuromodeling (the development of mathematical models for diagnostics of brain diseases) and their application to concrete clinical questions (Computational Psychiatry/Psychosomatics). The first part of the course will introduce disease concepts from psychiatry and psychosomatics, their history, and clinical priority problems. The second part of the course concerns computational modeling of neuronal and cognitive processes for clinical applications. A particular focus is on Bayesian methods and generative models, for example, dynamic causal models for inferring neuronal mechanisms from neuroimaging data, and hierarchical Bayesian models for inference on cognitive mechanisms from behavioural data. The course discusses the mathematical and statistical principles behind these models, illustrates their application to various psychiatric diseases, and outlines a general research strategy based on generative models. Lecture topics include: 1. Introduction to Translational Neuromodeling and Computational Psychiatry/Psychosomatics 2. Psychiatric nosology 3. Pathophysiology of psychiatric disease mechanisms 4. Principles of Bayesian inference and generative modeling 5. Variational Bayes (VB) 6. Bayesian model selection 7. Markov Chain Monte Carlo techniques (MCMC) 8. Bayesian frameworks for understanding psychiatric and psychosomatic diseases 9. Generative models of fMRI data 10. Generative models of electrophysiological data 11. Generative models of behavioural data 12. Computational concepts of schizophrenia, depression and autism 13. Model-based predictions about individual patients Practical exercises include mathematical derivations and the implementation of specific models or inference methods. In additional project work, students are required to use one of the examples discussed in the course as a basis for developing their own generative model and use it for simulations and/or inference in application to a clinical question. Group work (up to 3 students) is permitted. | |||||

Literature | See TNU website: https://www.tnu.ethz.ch/en/teaching.html | |||||

Prerequisites / Notice | Knowledge of principles of statistics, programming skills (MATLAB or Python) | |||||

227-1032-00L | Neuromorphic Engineering II Information for UZH students: Enrolment to this course unit only possible at ETH. No enrolment to module INI405 at UZH. Please mind the ETH enrolment deadlines for UZH students: Link | W | 6 credits | 5G | T. Delbrück, G. Indiveri, S.‑C. Liu | |

Abstract | This course teaches the basics of analog chip design and layout with an emphasis on neuromorphic circuits, which are introduced in the fall semester course "Neuromorphic Engineering I". | |||||

Objective | Design of a neuromorphic circuit for implementation with CMOS technology. | |||||

Content | This course teaches the basics of analog chip design and layout with an emphasis on neuromorphic circuits, which are introduced in the autumn semester course "Neuromorphic Engineering I". The principles of CMOS processing technology are presented. Using a set of inexpensive software tools for simulation, layout and verification, suitable for neuromorphic circuits, participants learn to simulate circuits on the transistor level and to make their layouts on the mask level. Important issues in the layout of neuromorphic circuits will be explained and illustrated with examples. In the latter part of the semester students simulate and layout a neuromorphic chip. Schematics of basic building blocks will be provided. The layout will then be fabricated and will be tested by students during the following fall semester. | |||||

Literature | S.-C. Liu et al.: Analog VLSI Circuits and Principles; software documentation. | |||||

Prerequisites / Notice | Prerequisites: Neuromorphic Engineering I strongly recommended | |||||

227-1034-00L | Computational Vision (University of Zurich) No enrolment to this course at ETH Zurich. Book the corresponding module directly at UZH. UZH Module Code: INI402 Mind the enrolment deadlines at UZH: https://www.uzh.ch/cmsssl/en/studies/application/mobilitaet.html | W | 6 credits | 2V + 1U | D. Kiper | |

Abstract | This course focuses on neural computations that underlie visual perception. We study how visual signals are processed in the retina, LGN and visual cortex. We study the morpholgy and functional architecture of cortical circuits responsible for pattern, motion, color, and three-dimensional vision. | |||||

Objective | This course considers the operation of circuits in the process of neural computations. The evolution of neural systems will be considered to demonstrate how neural structures and mechanisms are optimised for energy capture, transduction, transmission and representation of information. Canonical brain circuits will be described as models for the analysis of sensory information. The concept of receptive fields will be introduced and their role in coding spatial and temporal information will be considered. The constraints of the bandwidth of neural channels and the mechanisms of normalization by neural circuits will be discussed. The visual system will form the basis of case studies in the computation of form, depth, and motion. The role of multiple channels and collective computations for object recognition will be considered. Coordinate transformations of space and time by cortical and subcortical mechanisms will be analysed. The means by which sensory and motor systems are integrated to allow for adaptive behaviour will be considered. | |||||

Content | This course considers the operation of circuits in the process of neural computations. The evolution of neural systems will be considered to demonstrate how neural structures and mechanisms are optimised for energy capture, transduction, transmission and representation of information. Canonical brain circuits will be described as models for the analysis of sensory information. The concept of receptive fields will be introduced and their role in coding spatial and temporal information will be considered. The constraints of the bandwidth of neural channels and the mechanisms of normalization by neural circuits will be discussed. The visual system will form the basis of case studies in the computation of form, depth, and motion. The role of multiple channels and collective computations for object recognition will be considered. Coordinate transformations of space and time by cortical and subcortical mechanisms will be analysed. The means by which sensory and motor systems are integrated to allow for adaptive behaviour will be considered. | |||||

Literature | Books: (recommended references, not required) 1. An Introduction to Natural Computation, D. Ballard (Bradford Books, MIT Press) 1997. 2. The Handbook of Brain Theorie and Neural Networks, M. Arbib (editor), (MIT Press) 1995. | |||||

Statistics | ||||||

Number | Title | Type | ECTS | Hours | Lecturers | |

401-0102-00L | Applied Multivariate Statistics | W | 5 credits | 2V + 1U | F. Sigrist | |

Abstract | Multivariate statistics analyzes data on several random variables simultaneously. This course introduces the basic concepts and provides an overview of classical and modern methods of multivariate statistics including visualization, dimension reduction, supervised and unsupervised learning for multivariate data. An emphasis is on applications and solving problems with the statistical software R. | |||||

Objective | After the course, you are able to: - describe the various methods and the concepts behind them - identify adequate methods for a given statistical problem - use the statistical software R to efficiently apply these methods - interpret the output of these methods | |||||

Content | Visualization, multivariate outliers, the multivariate normal distribution, dimension reduction, principal component analysis, multidimensional scaling, factor analysis, cluster analysis, classification, multivariate tests and multiple testing | |||||

Lecture notes | None | |||||

Literature | 1) "An Introduction to Applied Multivariate Analysis with R" (2011) by Everitt and Hothorn 2) "An Introduction to Statistical Learning: With Applications in R" (2013) by Gareth, Witten, Hastie and Tibshirani Electronic versions (pdf) of both books can be downloaded for free from the ETH library. | |||||

Prerequisites / Notice | This course is targeted at students with a non-math background. Requirements: ========== 1) Introductory course in statistics (min: t-test, regression; ideal: conditional probability, multiple regression) 2) Good understanding of R (if you don't know R, it is recommended that you study chapters 1,2,3,4, and 5 of "Introductory Statistics with R" from Peter Dalgaard, which is freely available online from the ETH library) An alternative course with more emphasis on theory is 401-6102-00L "Multivariate Statistics" (only every second year). 401-0102-00L and 401-6102-00L are mutually exclusive. You can register for only one of these two courses. | |||||

401-3622-00L | RegressionDoes not take place this semester. | W | 8 credits | 4G | not available | |

Abstract | In regression, the dependency of a random response variable on other variables is examined. We consider the theory of linear regression with one or more covariates, high-dimensional linear models, nonlinear models and generalized linear models, robust methods, model choice and nonparametric models. Several numerical examples will illustrate the theory. | |||||

Objective | Introduction into theory and practice of a broad and popular area of statistics, from a modern viewpoint. | |||||

Content | In der Regression wird die Abhängigkeit einer beobachteten quantitativen Grösse von einer oder mehreren anderen (unter Berücksichtigung zufälliger Fehler) untersucht. Themen der Vorlesung sind: Einfache und multiple Regression, Theorie allgemeiner linearer Modelle, Hoch-dimensionale Modelle, Ausblick auf nichtlineare Modelle. Querverbindungen zur Varianzanalyse, Modellsuche, Residuenanalyse; Einblicke in Robuste Regression. Durchrechnung und Diskussion von Anwendungsbeispielen. | |||||

Lecture notes | Lecture notes | |||||

Prerequisites / Notice | Credits cannot be recognised for both courses 401-3622-00L Regression and 401-0649-00L Applied Statistical Regression in the Mathematics Bachelor and Master programmes (to be precise: one course in the Bachelor and the other course in the Master is also forbidden). | |||||

401-4632-15L | Causality | W | 4 credits | 2G | C. Heinze-Deml | |

Abstract | In statistics, we are used to search for the best predictors of some random variable. In many situations, however, we are interested in predicting a system's behavior under manipulations. For such an analysis, we require knowledge about the underlying causal structure of the system. In this course, we study concepts and theory behind causal inference. | |||||

Objective | After this course, you should be able to - understand the language and concepts of causal inference - know the assumptions under which one can infer causal relations from observational and/or interventional data - describe and apply different methods for causal structure learning - given data and a causal structure, derive causal effects and predictions of interventional experiments | |||||

Prerequisites / Notice | Prerequisites: basic knowledge of probability theory and regression | |||||

401-6102-00L | Multivariate Statistics | W | 4 credits | 2G | N. Meinshausen | |

Abstract | Multivariate Statistics deals with joint distributions of several random variables. This course introduces the basic concepts and provides an overview over classical and modern methods of multivariate statistics. We will consider the theory behind the methods as well as their applications. | |||||

Objective | After the course, you should be able to: - describe the various methods and the concepts and theory behind them - identify adequate methods for a given statistical problem - use the statistical software "R" to efficiently apply these methods - interpret the output of these methods | |||||

Content | Visualization / Principal component analysis / Multidimensional scaling / The multivariate Normal distribution / Factor analysis / Supervised learning / Cluster analysis | |||||

Lecture notes | None | |||||

Literature | The course will be based on class notes and books that are available electronically via the ETH library. | |||||

Prerequisites / Notice | Target audience: This course is the more theoretical version of "Applied Multivariate Statistics" (401-0102-00L) and is targeted at students with a math background. Prerequisite: A basic course in probability and statistics. Note: The courses 401-0102-00L and 401-6102-00L are mutually exclusive. You may register for at most one of these two course units. | |||||

401-6624-11L | Applied Time Series | W | 5 credits | 2V + 1U | M. Dettling | |

Abstract | The course starts with an introduction to time series analysis (examples, goal, mathematical notation). In the following, descriptive techniques, modeling and prediction as well as advanced topics will be covered. | |||||

Objective | Getting to know the mathematical properties of time series, as well as the requirements, descriptive techniques, models, advanced methods and software that are necessary such that the student can independently run an applied time series analysis. | |||||

Content | The course starts with an introduction to time series analysis that comprises of examples and goals. We continue with notation and descriptive analysis of time series. A major part of the course will be dedicated to modeling and forecasting of time series using the flexible class of ARMA models. More advanced topics that will be covered in the following are time series regression, state space models and spectral analysis. | |||||

Lecture notes | A script will be available. | |||||

Prerequisites / Notice | The course starts with an introduction to time series analysis that comprises of examples and goals. We continue with notation and descriptive analysis of time series. A major part of the course will be dedicated to modeling and forecasting of time series using the flexible class of ARMA models. More advanced topics that will be covered in the following are time series regression, state space models and spectral analysis. | |||||

Machine Learning and Artificial Intelligence | ||||||

Number | Title | Type | ECTS | Hours | Lecturers | |

252-0526-00L | Statistical Learning Theory | W | 7 credits | 3V + 2U + 1A | J. M. Buhmann | |

Abstract | The course covers advanced methods of statistical learning : Statistical learning theory;variational methods and optimization, e.g., maximum entropy techniques, information bottleneck, deterministic and simulated annealing; clustering for vectorial, histogram and relational data; model selection; graphical models. | |||||

Objective | The course surveys recent methods of statistical learning. The fundamentals of machine learning as presented in the course "Introduction to Machine Learning" are expanded and in particular, the theory of statistical learning is discussed. | |||||

Content | # Theory of estimators: How can we measure the quality of a statistical estimator? We already discussed bias and variance of estimators very briefly, but the interesting part is yet to come. # Variational methods and optimization: We consider optimization approaches for problems where the optimizer is a probability distribution. Concepts we will discuss in this context include: * Maximum Entropy * Information Bottleneck * Deterministic Annealing # Clustering: The problem of sorting data into groups without using training samples. This requires a definition of ``similarity'' between data points and adequate optimization procedures. # Model selection: We have already discussed how to fit a model to a data set in ML I, which usually involved adjusting model parameters for a given type of model. Model selection refers to the question of how complex the chosen model should be. As we already know, simple and complex models both have advantages and drawbacks alike. # Statistical physics models: approaches for large systems approximate optimization, which originate in the statistical physics (free energy minimization applied to spin glasses and other models); sampling methods based on these models | |||||

Lecture notes | A draft of a script will be provided; transparencies of the lectures will be made available. | |||||

Literature | Hastie, Tibshirani, Friedman: The Elements of Statistical Learning, Springer, 2001. L. Devroye, L. Gyorfi, and G. Lugosi: A probabilistic theory of pattern recognition. Springer, New York, 1996 | |||||

Prerequisites / Notice | Requirements: knowledge of the Machine Learning course basic knowledge of statistics, interest in statistical methods. It is recommended that Introduction to Machine Learning (ML I) is taken first; but with a little extra effort Statistical Learning Theory can be followed without the introductory course. | |||||

252-3005-00L | Natural Language Understanding Number of participants limited to 200. | W | 4 credits | 2V + 1U | M. Ciaramita, T. Hofmann | |

Abstract | This course presents topics in natural language processing with an emphasis on modern techniques, primarily focusing on statistical and deep learning approaches. The course provides an overview of the primary areas of research in language processing as well as a detailed exploration of the models and techniques used both in research and in commercial natural language systems. | |||||

Objective | The objective of the course is to learn the basic concepts in the statistical processing of natural languages. The course will be project-oriented so that the students can also gain hands-on experience with state-of-the-art tools and techniques. | |||||

Content | This course presents an introduction to general topics and techniques used in natural language processing today, primarily focusing on statistical approaches. The course provides an overview of the primary areas of research in language processing as well as a detailed exploration of the models and techniques used both in research and in commercial natural language systems. | |||||

Literature | Lectures will make use of textbooks such as the one by Jurafsky and Martin where appropriate, but will also make use of original research and survey papers. | |||||

261-5110-00L | Optimization for Data Science | W | 8 credits | 3V + 2U + 2A | B. Gärtner, D. Steurer | |

Abstract | This course teaches an overview of modern optimization methods, with applications in particular for machine learning and data science. | |||||

Objective | Understanding the theoretical and practical aspects of relevant optimization methods used in data science. Learning general paradigms to deal with optimization problems arising in data science. | |||||

Content | This course teaches an overview of modern optimization methods, with applications in particular for machine learning and data science. In the first part of the course, we will discuss how classical first and second order methods such as gradient descent and Newton's method can be adapated to scale to large datasets, in theory and in practice. We also cover some new algorithms and paradigms that have been developed specifically in the context of data science. The emphasis is not so much on the application of these methods (many of which are covered in other courses), but on understanding and analyzing the methods themselves. In the second part, we discuss convex programming relaxations as a powerful and versatile paradigm for designing efficient algorithms to solve computational problems arising in data science. We will learn about this paradigm and develop a unified perspective on it through the lens of the sum-of-squares semidefinite programming hierarchy. As applications, we are discussing non-negative matrix factorization, compressed sensing and sparse linear regression, matrix completion and phase retrieval, as well as robust estimation. | |||||

Prerequisites / Notice | As background, we require material taught in the course "252-0209-00L Algorithms, Probability, and Computing". It is not necessary that participants have actually taken the course, but they should be prepared to catch up if necessary. | |||||

263-3710-00L | Machine Perception Number of participants limited to 150. | W | 5 credits | 2V + 1U + 1A | O. Hilliges | |

Abstract | Recent developments in neural networks (aka “deep learning”) have drastically advanced the performance of machine perception systems in a variety of areas including computer vision, robotics, and intelligent UIs. This course is a deep dive into deep learning algorithms and architectures with applications to a variety of perceptual tasks. | |||||

Objective | Students will learn about fundamental aspects of modern deep learning approaches for perception. Students will learn to implement, train and debug their own neural networks and gain a detailed understanding of cutting-edge research in learning-based computer vision, robotics and HCI. The final project assignment will involve training a complex neural network architecture and applying it on a real-world dataset of human activity. The core competency acquired through this course is a solid foundation in deep-learning algorithms to process and interpret human input into computing systems. In particular, students should be able to develop systems that deal with the problem of recognizing people in images, detecting and describing body parts, inferring their spatial configuration, performing action/gesture recognition from still images or image sequences, also considering multi-modal data, among others. | |||||

Content | We will focus on teaching: how to set up the problem of machine perception, the learning algorithms, network architectures and advanced deep learning concepts in particular probabilistic deep learning models The course covers the following main areas: I) Foundations of deep-learning. II) Probabilistic deep-learning for generative modelling of data (latent variable models, generative adversarial networks and auto-regressive models). III) Deep learning in computer vision, human-computer interaction and robotics. Specific topics include: I) Deep learning basics: a) Neural Networks and training (i.e., backpropagation) b) Feedforward Networks c) Timeseries modelling (RNN, GRU, LSTM) d) Convolutional Neural Networks for classification II) Probabilistic Deep Learning: a) Latent variable models (VAEs) b) Generative adversarial networks (GANs) c) Autoregressive models (PixelCNN, PixelRNN, TCNs) III) Deep Learning techniques for machine perception: a) Fully Convolutional architectures for dense per-pixel tasks (i.e., instance segmentation) b) Pose estimation and other tasks involving human activity c) Deep reinforcement learning IV) Case studies from research in computer vision, HCI, robotics and signal processing | |||||

Literature | Deep Learning Book by Ian Goodfellow and Yoshua Bengio | |||||

Prerequisites / Notice | This is an advanced grad-level course that requires a background in machine learning. Students are expected to have a solid mathematical foundation, in particular in linear algebra, multivariate calculus, and probability. The course will focus on state-of-the-art research in deep-learning and will not repeat basics of machine learning Please take note of the following conditions: 1) The number of participants is limited to 150 students (MSc and PhDs). 2) Students must have taken the exam in Machine Learning (252-0535-00) or have acquired equivalent knowledge 3) All practical exercises will require basic knowledge of Python and will use libraries such as TensorFlow, scikit-learn and scikit-image. We will provide introductions to TensorFlow and other libraries that are needed but will not provide introductions to basic programming or Python. The following courses are strongly recommended as prerequisite: * "Visual Computing" or "Computer Vision" The course will be assessed by a final written examination in English. No course materials or electronic devices can be used during the examination. Note that the examination will be based on the contents of the lectures, the associated reading materials and the exercises. | |||||

Big Data Systems | ||||||

Number | Title | Type | ECTS | Hours | Lecturers | |

252-3900-00L | Big Data for Engineers This course is not intended for Computer Science and Data Science students! | W | 6 credits | 2V + 2U + 1A | G. Fourny | |

Abstract | This course is part of the series of database lectures offered to all ETH departments, together with Information Systems for Engineers. It introduces the most recent advances in the database field: how do we scale storage and querying to Petabytes of data, with trillions of records? How do we deal with heterogeneous data sets? How do we deal with alternate data shapes like trees and graphs? | |||||

Objective | The key challenge of the information society is to turn data into information, information into knowledge, knowledge into value. This has become increasingly complex. Data comes in larger volumes, diverse shapes, from different sources. Data is more heterogeneous and less structured than forty years ago. Nevertheless, it still needs to be processed fast, with support for complex operations. This combination of requirements, together with the technologies that have emerged in order to address them, is typically referred to as "Big Data." This revolution has led to a completely new way to do business, e.g., develop new products and business models, but also to do science -- which is sometimes referred to as data-driven science or the "fourth paradigm". Unfortunately, the quantity of data produced and available -- now in the Zettabyte range (that's 21 zeros) per year -- keeps growing faster than our ability to process it. Hence, new architectures and approaches for processing it were and are still needed. Harnessing them must involve a deep understanding of data not only in the large, but also in the small. The field of databases evolves at a fast pace. In order to be prepared, to the extent possible, to the (r)evolutions that will take place in the next few decades, the emphasis of the lecture will be on the paradigms and core design ideas, while today's technologies will serve as supporting illustrations thereof. After visiting this lecture, you should have gained an overview and understanding of the Big Data landscape, which is the basis on which one can make informed decisions, i.e., pick and orchestrate the relevant technologies together for addressing each business use case efficiently and consistently. | |||||

Content | This course gives an overview of database technologies and of the most important database design principles that lay the foundations of the Big Data universe. It targets specifically students with a scientific or Engineering, but not Computer Science, background. The material is organized along three axes: data in the large, data in the small, data in the very small. A broad range of aspects is covered with a focus on how they fit all together in the big picture of the Big Data ecosystem. - physical storage (HDFS, S3) - logical storage (key-value stores, document stores, column stores, key-value stores, data warehouses) - data formats and syntaxes (XML, JSON, CSV) - data shapes and models (tables, trees, graphs) - an overview of programming languages with a focus on their type systems (SQL, XQuery) - the most important query paradigms (selection, projection, joining, grouping, ordering, windowing) - paradigms for parallel processing (MapReduce) and technologies (Hadoop, Spark) - optimization techniques (functional and declarative paradigms, query plans, rewrites, indexing) - applications. Large scale analytics and machine learning are outside of the scope of this course. | |||||

Literature | Papers from scientific conferences and journals. References will be given as part of the course material during the semester. | |||||

Prerequisites / Notice | This course is not intended for Computer Science and Data Science students. Computer Science and Data Science students interested in Big Data MUST attend the Master's level Big Data lecture, offered in Fall. Requirements: programming knowledge (Java, C++, Python, PHP, ...) as well as basic knowledge on databases (SQL). If you have already built your own website with a backend SQL database, this is perfect. Attendance is especially recommended to those who attended Information Systems for Engineers last Fall, which introduced the "good old databases of the 1970s" (SQL, tables and cubes). However, this is not a strict requirement, and it is also possible to take the lectures in reverse order. |

- Page 1 of 2 All