As a result of today's meeting what is agreed is actually to continue on the light reading group format with some small modifications:
1. We will have a reading group every two weeks
2. We will have a pool of papers to read and we will select one paper from that pool for every reading group. (To initiate this pool of papers, please write your paper suggestions as a comment to this post - send at least one paper suggestion until February 5)
3. Everybody will make their best to come to the reading group by reading the paper.
4. We will continue to the blog posts and google calendar entries for communication. You can setup your google calendar so that you receive reminders for every new entry.
So basically nothing has changed with respect to the last year except than the pool of papers. For me, last year's light reading groups were not that effective. If at one point we feel the same, then we will discuss again.
Another thing, at the meeting at some point it is said that "if nobody likes the paper the presenter will change it". I now think that maybe this is not a good idea. We should count on each others decision and read the paper anyway.
The ones who would like to have a heavy reading group will make a separate meeting.
Lastly, the first reading group will be on 12 February. Along with your paper suggestions please also say if you would like to be the first one.
oya
Subscribe to:
Post Comments (Atom)
Posts
You can copy the papers you want to put in the pool to the following folder:
ReplyDelete/idiap/group/socialcomputing/lightReadingGroup_PaperPool
everybody in the "socialcomputing" group has reading/writing rights to this folder.
If you are not in the group, let me know
oya
I copied the following files to the paper pool:
ReplyDeletelaptev05periodic.pdf:
Periodic Motion Detection and Segmentation via
Approximate Sequence Alignment, Laptev et al., ICCV, 2005
lepri09modeling.pdf:
Modeling the Personality of Participants During Group Interactions, Lepri et al., UMAP, 2009
otsuka07automatic.pdf:
Automatic Inference of Cross-modal Nonverbal
Interactions in Multiparty Conversations, Otsuka et al., ICMI, 2007
I won't be able to present for the first reading group, since I have deadlines for mid february
oya
I would like to do discriminative vs generative classifiers. There are 2 choices for this: (a) A paper by Ng, Jordon (b) a chapter by mitchell. But if you feel these papers are a bit heavy, i have some easier choices on (c) significan testing (d) theory of measurements. The 4 files are in /idiap/group/socialcomputing/lightReadingGroup_PaperPool as
ReplyDelete(a) hari_ng.ps
(b) hari_mitchell.pdf
(c) hari_performance.pdf
(d) hari_measurement.pdf
i am a bit busy now; so i would like to do this sometime in march...
hari
I've copied a paper called Chang-ReadingTeaLeavesHowHumansInterpretTopicModels.pdf, from Blei's group. It is light enough, although it's been published in NIPS this year. I do not mind going first if you're all busy.
ReplyDeleteI have copied a paper on 'Deep Learning by Geoff Hinton'. I can go after Radu.
ReplyDeleteI've copied a paper called "Structure and Tie Strengths in Mobile Communication Networks". I can present after Dinesh :)
ReplyDeleteI propose an applicative paper titled "Reducing the Dimensionality of Data with Neural Networks". I can present in April.
ReplyDeletepath: /idiap/group/socialcomputing/lightReadingGroup_PaperPool/ReducingTheDimensionalityOfDataWithNeuralNetworks_hinton.pdf
I copied the following files to the pool,
ReplyDeleteZhang&Zhang:CVPR04 Hidden semantic concept discovery.pdf:
(CBIR) based on developing a hidden semantic concept
discovery methodology.
Bharat:SDM08 On the Dangers of Cross-Validation.pdf:
the risk for overfitting increases and the performance estimated by cross validation is no longer an effective estimate of generalization under large number of models.
Dueck&Frey:ICCV07: Non-metric affinity propagation for unsupervised image categorization.pdf:
23 pp, but just need to read the first 5. Clever clustering that finds the optimal number of clusters by itself.
Paco.
I propose:
ReplyDelete-Combining Collective Classification and Link Prediction
- Collective vs Independent Classification in Statistical Relational Learning
I can present either march or april.
Dayra