2017-18
| Nombre del curso: | Intelligent systems architecture |
|---|---|
| Course Name: | Intelligent systems architecture |
| Créditos: | 4 |
| Profesor: | María de Arteaga – Carnegie Mellon University |
| Hernán Astudillo Ph.D. – Universidad Técnica Federico Santa María – Chile | |
| Artur Dubrawski – Carnegie Mellon University | |
| Álvaro Riascos Ph.D – Universidad de los Andes |
| Programa del DISC: | Válido por: |
|---|---|
| MATI | Curso de profundización |
| Otras maestrías | Curso electivo |
| Pregrado | Electiva profesional |
Description
Modern enterprise information systems involve the use of structured and unstructured data to support decision making, and optimize business processes. To tackle these new challenges, it is necessary to apply novel analytic and technological mechanisms to support the design, build, test, and deployment of intelligent systems.
In this context, IT architects face new challenges. First, they need to interact with new stakeholders (i.e. data scientists) to understand architectural requirements related to big data ecosystems, data intensive algorithms and scalable architectures to achieve robust, flexible, useful and cost-effective solutions.
This course presents three complementary modules, relevant to modern IT architects. First, intelligent systems must be designed and developed following new architectural approaches. Experimental software engineering is also presented as a mechanism to validate analytical solutions. Second, foundations of statistical learning are presented to understand the main tasks and requirements faced by data scientists. Third, advanced analytics concepts and real world problems in the domain of intelligent systems are presented. In this module students will apply the concepts learned in the previous two modules to propose architectures designed to handle large scale data applications.
At the end of this course, you will be able to:
- Summarize the features and value of experimental software engineering and software architecture in analytical projects
- Define and validate non-functional requirements in the context of Intelligent Systems architectures
- Describe the foundations of machine learning and its uses in Intelligent Systems solutions.
- Analyze and compare different scenarios of applied data science in Intelligent Systems through pragmatic case studies.
- Integrate acquired knowledge to architecting Intelligent Systems
I – Experimental Software Engineering and Architecture Design
Prof. Hernán Astudillo
Session 1 (6/12/2017):
- Software architecture: motivation, process, description, quality, reference architectures, ecosystems.
- Architecture process: drivers, core ideas, structure, traceability.
- Case study: SNAM v1 as-is/to-be.
Session 2 (6/13/2017):
- Architecture elaboration: knowledge reuse – components, frameworks, styles, patterns, tactics.
- Knowledge construction: experimental software engineering, case studies, systematic reviews, experiments.
- Architectural knowledge: decisions capture, presentation, sharing, reuse.
- Case: PUA designers decision process (using DVIA).
- Case study: SNAM v2 to-be (reusing knowledge).
Session 3 (6/14/2017):
- Quality attributes (non-functional requirements): performance, availability, scalability.
- Architecture evaluation: quality models, trade-offs, softgoals, SAAM, ATAM.
- Case study: SNAM v3 to-be (quality-evaluated).
- Case: ExoBanco.
Session 4 (6/15/2017):
- Case study: MySpace – drivers, alternatives, trade-offs.
- Case study: big data architectures – drivers, alternatives, trade-offs, reference architectures.
II – Fundamentals of Machine Learning
Prof. Alvaro Riascos
Session 1 (6/20/2017):
- Introduction to statistical learning: risk, approximation and estimation error, overfitting, optimal regression and classification algorithms, bounds, consistency, no free lunch theorems. References: [JWHT], Chapter 1. [HTF], Chapter 1. [LS]. [JWHT], Sections 2.1, 2.2.
- Linear methods of regression and classification and regularization. References:[JWHT], Chapter 3,4. [HTF], Chapter 3,4.
Session 2 (6/21/2017):
- Non-linear methods and regularization. [HTF]: Chapter 5,6.
- Trees, boosting and random forests. [HTF]: Chapter 9,10,15.
Session 3 (6/22/2017):
- Neural networks and an introduction to deep learning. [B]: Chapter 5. [DLB]: Selected topics.
III- Intelligent Systems
Profs. Artur Dubrawski , María De Arteaga
Session 1 (6/28/2017):
- Engineering and characterization of data: Bringing data to shape for analysis. Clustering, characterizing, and summarizing data.
Session 2 (6/29/2017):
- Modeling expectations. Detecting anomalies and anomalous patterns.
- Case studies: surveillance of public health, food safety, breaks of billing patterns in health insurance.
Session 3 (6/30/2017):
- Learning from multi-modal data: How to learn from text, images, videos, and sounds.
Session 4 (7/4/2017):
- Graphical models: Probabilities, Bayesian statistics, graphical models.
Session 5 (7/5/2017):
- Learning with scarce labels: Active learning, semi-supervised learning, transfer learning.
- How things can go wrong: Non-iid data, correlation is not causation, unfair algorithms.