Notes for week 11

Barbara

The computational questions I want to have people discuss that arise from these are the following:

We've already discussed some models and evidence for classical conditioning (very generally speaking) in the basal ganglia, nucleus accumbens and amygdala. Here we have a situation where a particular class of information is privileged in some animals because of the presence of vasopressin/oxytocin receptor in the nucleus where vasopressin/oxytocin and dopamine converge.

I want to work out how this model might be realized a little more.

• What are the inputs doing there if there aren't receptors for them in the non-monogamous?
• How specific does the information have to be?

Then:

• how to code a fearsome US (in various sensory domains)?
• how to privilege or otherwise distinguish it from general sensory info?

Shimon

Zald

Some questions:

• What is the history of the amygdala = emotional processing notion?
• What is the location of the fabled "pleasure center" in the rat? Is it the nucleus accumbens?
• In the abstract, Zald mentions 10 "hypotheses" about the function of the amygdala in humans —
  1. the amygdala activates during exposure to aversive stimuli from multiple sensory modalities;
  2. the amygdala responds to positively valenced stimuli, but these responses are less consistent than those induced by aversive stimuli;
  3. amygdala responses are modulated by the arousal level, hedonic strength or current motivational value of stimuli;
  4. amygdala responses are subject to rapid habituation;
  5. the temporal characteristics of amygdala responses vary across stimulus categories and subject populations;
  6. emotionally valenced stimuli need not reach conscious awareness to engage amygdala processing;
  7. conscious hedonic appraisals do not require amygdala activation;
  8. activation of the amygdala is associated with modulation of motor readiness, autonomic functions, and cognitive processes including attention and memory;
  9. amygdala activations do not conform to traditional models of the lateralization of emotion;
  10. the extent and laterality of amygdala activations are related to factors including psychiatric status, gender and personality.
  Most of these, however, are negative, equivocal, or qualified/questioned in the text. Is there a better summary?

Young & Wang

• Looks like the mechanisms outlined by Young (joint action of oxytocin/vasopressin and dopamine) can explain cross-species attachment, or perhaps even falling in love with one's notebook computer.

Goodson

• Fig.1: Why present the Newman social behavior network as a 2D surface instead of six separate bars, or an actual six-node network?

  BTW: there is plenty of room in a six-dimensional space; thus, many different behaviors can be encoded by points in such a space (a propos this: high-dimensional spaces are counterintuitive in several ways).

• Fig.2C: Is there anything special about arginine vasotocin (AVT) that makes it (and its derivatives) particularly suitable for mediating the physiological functions it originally mediated? (I assume that with respect to the behavioral control that AVT derivatives were later co-opted to mediate this question no longer makes sense.)
• Fig.4: Looks like the sheer quantity of the relevant neuropeptide controls the intensity of the corresponding behavior(s). This seems to fit a general pattern of division of labor between pharma and neuro control pathways, the former regulating the intensity, but not modifying the structure, of the latter.
• Fig.5: What is immediate early gene response and what genes are Fos and Zenk?

Oxytocin and vasopressin

Computational remarks/questions/issues:

• In diversifying into a new ecological niche, it seems easier to adapt the neural architecture to the existing pharmacology than the other way around.
• It seems even easier to make do with the existing gross neural architecture and pharmacology, by sticking to a proven repertoire of behaviors.

• Speaking about behaviors, to what extent is territoriality the same in male midshipman fish and lions (a propos vasopressin)? And what is the female midshipman counterpart of a nipple (a propos isotocin/oxytocin)?

Encoding "privileged" information

• Neural encoding of the concept of nest in the mouse brain, L. Lin, G. Chen, H. Kuang, D. Wang, and J. Z. Tsien, PNAS 104:6066-6071 (April 3, 2007).

[probably for next time (week 12)]

One-trial learning

• FPGA
• Hebb with a very high value of the learning constant (as in Unsupervised learning of visual structure, Edelman et al. (2002)
• Bayesian one-shot learning (Fei-Fei, Fergus, and Perona, 2003)
• See also the Wiki article