ME 461/561-012: Speical Topic
Estimation and Learning for Autonomous Systems
| Sections | Time | Location |
|---|---|---|
| ME 461-012 & ME 561-012 | Tue & Thu 12:30-13:45 | Mechanical Engineering 210 |
Quick links
Syllabus
Class Time & Space: Tue & Thu 12:30 - 13:45 @ME 210
Course Instructor:
Wenbin Wan; email: wwan@unm.edu; office hour: Thu 11am-noon @ME 425
Prerequisites: ME380 and MATH 316 or equivalent, concurrent registration in ME480 preferred; or consent of instructor
Textbook (optional): Barfoot, T. D., State estimation for robotics. Cambridge University Press, 2017.
Learning Objectives:
- Introduction to key concepts in state estimation and machine learning
- Gain basic familiarity with the CPS topics
- Control/Estimation algorithms
- Computational models
- Foundational connections between control/estimation and machine learning
- Exposure to some most influential ideas in the community
Tentative Topics:
- Mathematical Background
- Introduction to optimization
- Matrix algebra and matrix calculus
- Linear systems theory
- Probability theory
- State estimation
- Least squares estimation
- Propagation of states and covariances
- Discrete-time Kalman Filter
- Alternate Kalman Filter formulations
- Verifying Kalman Filter performance
- Robust Kalman Filtering
- Nonlinear Kalman filtering
- Unscented Kalman filter
- Particle filter
- Inference and learning
- Introduction to machine learning
- Neural networks
- Stochastic approximation
- Stochastic gradient descent
- Monte Carlo methods
- Importance sampling
- Parameter estimation via EM algorithm
- Variational inference
- Variational lower bound (ELBO)
- Auto-Encoding Variational Bayes (VAE)
- Monte Carlo gradient estimation methods in machine learning
Course Assessments:
| Assessment | Credit | Comment |
|---|---|---|
| Homeworks | 50% | 3 writing homeworks and 2 mini coding projects |
| Participation/attendance | 5% | 2 unexcused absences |
| Final project | 45% | Proposal (15%) + Final report (30%) |
| Extra credits | 5% | LaTex’d assignments - 1% each (up to 5%) |
Grading:
The total percentage \(p\)% corresponds the final grades as follows.
- A+, if \(p \in\) [98,105]
- A, if \(p \in\) [92,98)
- A-, if \(p \in\) [90,92)
- B+, if \(p \in\) [88,90)
- B, if \(p \in\) [82,88)
- B-, if \(p \in\) [70,82)
- C+, if \(p \in\) [68,70)
- C, if \(p \in\) [62,68)
- C-, if \(p \in\) [60,62)
- F, if \(p \in\) [0,60)
Course Policies:
Ethics
- You are expected to abide by the University policies on academic honesty and integrity as given in the Student Handbook. Violations of these policies will not be tolerated and are subject to severe sanctions up to and including expulsion from the university.
Behaviors
- Please show respect for your classmates by limiting distractive behavior. You may use a smart phone, labtop or tablet to take notes or for other academic purposes directly related to the class. Mute your cell phones and other devices, and please keep any side discussions short and quiet.
Work Habits
- Due dates are non-negotiable, and late work will NOT be accepted!
Other Policies
- In accordance with university policies, I will make reasonable accommodation to a student’s religious observances and practices due to national origin. If you must miss class because of a feast day or religious holiday, please inform me promptly.
- UNM is committed to providing equitable access to learning opportunities for students with documented disabilities. As your instructor, it is my objective to facilitate an inclusive classroom setting, in which students have full access and opportunity to participate. To engage in a confidential conversation about the process for requesting reasonable accommodations for this class and/or program, please contact Accessibility Resource Center at aresrvs@unm.edu or by phone at 505-277-3506.
Schedule
| Dates | Topics | Materials | Comments |
|---|---|---|---|
| Week 1 (8/22 8/24) |
Syllabus reading & Introduction Optimization basics Optimization in \(\mathbb{R}^n\) |
[00.pdf] [01.pdf] [02.pdf] |
|
| Week 2 (8/29 8/31) |
Convex optimization Linear systems |
[03.pdf] [04.pdf] |
|
| Week 3 (9/05 9/07) |
Probability review Least squares estimation |
[05.pdf] [06.pdf] |
|
| Week 4 (9/12 9/14) |
Propagation of states and covariances Discrete-time Kalman filter |
[07.pdf] [08.pdf] |
HW1 Due @Sat. 11pm |
| Week 5 (9/19 9/21) |
Alternate Kalman filter formulations Constrained Kalman filter |
[09.pdf] [10.pdf] |
|
| Week 6 (9/26 9/28) |
Verifying Kalman filter performance Robust Kalman filtering |
[11.pdf] [12.pdf] |
HW2 Due @Sat. 11pm |
| Week 7 (10/03 10/05) |
\(\mathcal{H}_{\infty}\) filter Machine learning basics |
[13.pdf] [14.pdf] |
|
| Week 8 (10/10 |
Regression methods | [15.pdf] | 10/12: No class - Fall break |
| Week 9 (10/17 10/19) |
Neural networks I Neural networks II |
[16.pdf] [17.pdf] |
HW3 Due @Sat. 11pm |
| Week 10 (10/24 10/26) |
Support vector machines Gaussian processes |
[18.pdf] [19.pdf] |
Initialize your project proposal |
| Week 11 (10/31 11/04) |
Gaussian process regression Bayesian recursive estimation using particle filter |
[20.pdf] [21.pdf] |
HW4 Due @Sat. 11pm |
| Week 12 (11/07 11/09) |
Monte Carlo Integration & Importance sampling Bootstrap |
[22.pdf] [23.pdf] |
|
| Week 13 (11/14 11/16) |
Parameter estimation via EM algorithm & Variational lower bound | [24.pdf] | 11/16: Project proposal discussion |
| Week 14 (11/21 |
Auto-encoding variational Bayes | [25.pdf] | 11/23: No class - Thanksgiving |
| Week 15 (11/28 11/30) |
Hidden Markov models Advanced topics |
[26.pdf] [27.pdf] |
|
| Week 16 (12/05 12/07) |
Final project presentation HW5 Due @Sat. 11pm |
||
| Final Exam Week (12/11-12/16) |
Final report Due |
Page maintained by Wenbin Wan