[TNJC] Neuro Journal Club this week (14/3, 2pm, Friday)

6 views
Skip to first unread message

Ben Dongsung Huh

unread,
Mar 13, 2014, 2:45:35 PM3/13/14
to gatsby-tn...@googlegroups.com

Dear all,
We'll have a Cosyne wrap-up session for this week's TNJC.
Each attendee presents their favorite talk/poster from the meeting.
I created a google doc so please write down your preferred talk/poster to avoid overlap.

See you then!


Next presenter list:

Interview Week  3/21/2014
Vincent            3/28/2014
Kristin              4/4/2014
Peter L             4/11/2014


Ben Dongsung Huh
Gatsby Computational Neuroscience Unit

Agnieszka Grabska Barwinska

unread,
Mar 13, 2014, 8:15:28 PM3/13/14
to Ben Dongsung Huh, gatsby-tn...@googlegroups.com
Dear All,

I'm sorry to miss another JC, looking forward to it moving back to Tuesdays..

Since I won't be there to talk about my impressions from Cosyne (great
snow on Sunday! ;)), I wrote down the talks/posters that made the
greatest impression on me.

Doris Tsao's talk on face processing system - it's been a while since
I last heard about FFA. Lots of progress made since then, very
impressive and convincing arguments for the (now six patches of ) area
to be specialised in coding for faces. Not very computational, but
still, I'd say, most interesting of the talks I've seen. If no one
covers it, I'll probably steal the topic for a tea talk next week.

Nachum Ulanovsky's talk on grid cells and place cells in bats - again,
nothing computational, but a strong evidence against theta-gamma
oscillations being of any relevance to coding of space by hippocampus
and the rest..

Neir Eshel gave an interesting account of (ventral tegmental area)
dopamine neurons responses in face of rewards and their predictions
(signalled by odours). They recorded both divisive effect of
prediction on the population of neurons, and "scaled" subtractive
inhibition on single neurons, where "scaling" was predicted by the
responsiveness of neurons to rewards (i.e. the more they fired to the
reward, the more would be subtracted from the activity, when the
reward could be predicted with 100% certainty).

My pick of the poster session was "An olfactory cocktail party -
Detection of single odorants within mixtures in rodents" (from V
Murthy's lab) - a title you might recognise from my own posters.
Finally, some experimental confirmation that we're not solving an
abstract task! ;)

Most memorable workshop sessions I've seen were by Peter L (you'll
probably hear about that one) and Jonathan Pillow (my other pick for
the tea talk, if no one else covers that one).

Looking forward to hearing back what you discussed,

a
> --
> You received this message because you are subscribed to the Google Groups
> "Gatsby TNJC discussion group" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to gatsby-tnjc-dis...@googlegroups.com.
> Visit this group at http://groups.google.com/group/gatsby-tnjc-discuss.

Joana Soldado Magraner

unread,
Mar 14, 2014, 4:47:22 PM3/14/14
to Agnieszka Grabska Barwinska, Ben Dongsung Huh, gatsby-tn...@googlegroups.com
Hi all,

I am also sorry that I missed the discussion today, so I will briefly comment on some of the talks I found interesting.


The thing that surprised me the most was the finding by Yates et al. regarding the functional role of LIP in primate cortex. Surprisingly, LIP inactivation during a decision-making task did not impair performance, suggesting that LIP is not involved in the decision. Their results conflicted with the well-established belief that LIP main task is to integrate sensory evidence from area MT. The authors speculate that LIP may be accumulating post-decisional signals or that it may be read out in a very different manner than MT.


I was told that this talk was already mentioned during the discussion, so I will comment on something else, this time about one of the workshops: Altered Dynamics of Neural Circuit Activity in Brain Disorders.

The title of the talk was Neural dynamics in models of human focal epilepsy, by T. Sejnowski. He started the talk pointing out that epileptic patients are in fact a unique 'preparation'. He literally said: what better than a human for a human?
He presented a model based on delay differential equations (DDEs). These relate the derivative of a signal to the signal itself but shifted in time. The equations can incorporate several delays and non-linearities. The goal was to capture the underlying dynamics of EEG signals, coming from different patients and brain areas. Even though there are many possible combinations of delays and non-linearities, they could fit the data of 15 subjects and different brain areas using only 4 models. It would be interesting to find whether there's actually a biophysical correspondence to these delays and non-linear terms.
Finally, he announced that they are organizing a competition for seizure prediction algorithms. They are generating a big data base with recordings from many patients, which will be publicly available. Of course, there's going to be a substantial prize for the winner, hoping that this will be attractive enough to the machine learning community!



And yes, the Noise Correlation Workshop... that was fun indeed :P



Have a nice weekend!




Joana Soldado Magraner
Reply all
Reply to author
Forward
0 new messages