6.S064: Introduction to Machine Learning

A rough grade distribution:

A+ & A: 48% (grade > 90)
B: 41% (90 > grade > 74)
C: 11% (74 > grade > 40)
D: a few
#pin

For the last project we will offer a penalized acceptance of late submissions. In other words, if you can no longer use either of your two week-long free extensions, or do not have a note from the dean, then you will be penalized 5% for each additional 24h increments that you need prior to submission.
#pin

The teaching staff has posted a new lecture notes resource.

Title: RLRobotsMIT2013.pdf
http://www.piazza.com/class_profile/get_resource/hckx35ial496d5/hgfgxjymvjm553


links to the videos:

AIBO gait optimization (Kohl and Stone):
http://www.cs.utexas.edu/~AustinVilla/?p=research/learned_walk

Ball-in-Cup (Kober et al.):
http://www.youtube.com/watch?v=5oBAYbOF2Qo&list=UULA72CHSZ11x1xZk40cazLQ&index=14

Robot table-tennis (Muelling et al.):
http://www.youtube.com/watch?v=SH3bADiB7uQ&list=UULA72CHSZ11x1xZk40cazLQ&index=1

The Punchulum (Kuindersma et al.):
http://www.youtube.com/watch?feature=player_embedded&v=gYLqb7t97ik

The Stanford helicopter (Abbeel et al.):
http://www.youtube.com/watch?v=VCdxqn0fcnE

Cart-Pole Swing Up (Deisenroth and Rasmussen):
http://www.youtube.com/watch?v=XiigTGKZfks

Autonomous Robot Skill Acquisition (Konidaris et al.):
http://www.youtube.com/watch?v=yUICAkSQTZY

Hi,

You are granted an extension for project 3 until Sunday.

For those of you who want to use your free week extension, you can apply the, for both the homework 9 and the project. #project3
#pin

Project 2 grade is out. You can access to it on the Gradebook Module of stellar website (https://stellar.mit.edu/S/course/6/sp13/6.S064/index.html).

Please note that we didn't grade those we don't receive any submission. So if you submitted before but don't have the grades, please contact TA and send your submission by email.

A rough grading rubic for project 2:

Task 1: PCA and Image reconstruction: 25pt (cut off 5 if missing some figures)

Task 2: Classification:
(a) 7pt each curve (4*7 = 28)
(b) 5 points for each comment (two comment in total) (5*2 = 10)

Task 3: Clustering:
(1) 5pt for a reasonable result; 5pt for giving results of all the trials; 5pt for
analysis (15)

(2) 7 for the plot; 5 for discussion (12)

(3) 2.5 point for each finding (5 in total); 5 for analysis (10)

100 points in total.
#pin

The computation of log-likelihood in the previous code is incorrect. We fix it so that now the log-likelihoods for each model are non-decreasing. The code and instruction are updated. You can download it in @440.

If you have already done implementation in the previous framework, you can simply copy your code to the same place in the new version and test again. We only change the computation of log-likelihood, so it shouldn't affect your implementation.
#pin

• Project 3 (the last one) is released. You can download the zip pack in the attachment and all the files needed (instructions, provided code, data etc.) can be found in it.

• Deadline: May 10th.

• Sorry for the delayed release and close deadline. You won't have to write a lot of code this time, but it may take efforts to get started. We will help you with this project on tomorrow's recitation. It will be helpful.

• About the submission: please submit the completed code and a PDF report with your result and analysis. More submission details will come out later.
  
  
• Correction: the computation of log-likelihood in the previous code is incorrect. We fix it so that now the log-likelihoods for each model are non-decreasing. The code and instruction are updated.

#pin