Fwd: [Comp-neuro] From firing rate analysis to "concept cells" dogma

Begin forwarded message:

From: Dorian Aur <dorianaur@gmail.com>
Date: August 23, 2011 2:11:03 PM
To: Connectionists@cs.cmu.edu,cogsci@psych.colorado.edu,agents@cs.umbc.edu,ai@dcs.qmul.ac.uk,aiia_announcements@penelope.csd.auth.gr,COGPSY@listservtamu.edu,comp-neuro@neuroinf.org,spp-misc@philebus.tamu.edu,comp-neuro-owner@neuroinf.org
Subject: [Comp-neuro] From firing rate analysis to "concept cells" dogma

I have received many letters from friends, colleagues that asked me to summarize the entire debate posted on that "private" short list. Since it is our general interest to understand the general issues regarding firing rate and the myth of temporal coding the comments posted by Asim Roy, Christof Koch, Quilan Quiroga  and Asher Evans are revealed. These comments are relevant to understand the entire scientific debate, the issues regarding the concept cell idea. After 6 days of discussion several members on this short - list understood that there are serious issues regarding "concept cells" (or computation in single neurons).  As you may see the entire discussion was about "High level abstractions in concept cells; Single spiking neurons have meaning and are actually at the cognitive level"

Details regarding KEY moments of the debate:

1.      Aug 12: I was invited on a "private" list after the message was published on comp-neuro and on their broad list .  Asim Roy read my message and acknowledges that may be some issues regarding the firing rate measure and that may be other "better methods"
2.      Aug 15: I have informed all members of the "private list" about semantics in single spike, spike directivity and  two counterexamples were presented.
3.       Aug 16:  I posted  other two other examples regarding neurofeedback to show  that Moran Cerf paper is not about "concept cells".
4.      Aug 17: Koch is forced to provide explanations and identifies another paper  (Quiroga et al. 2007) where the same group found so called "concept cells". It seems that this lab had a history of "breakthrough findings". Koch's comments are questioned, other members start to make inquiries regarding the "concept cell" idea
5.      Aug 18:Asim Roy, cannot answer to basic questions -definition of concept cell, grandmother cell: Asim Roy:"I have characterized it in a recent paper. But I don't know of an exact definition." Quian Quiroga tries to help, presents the story of grandmother cell, not the expected definition

6.      AUG 19: I HAVE CHANGED THE FROM MORAN CERF TOPIC AND POINT TO THE REAL ISSUE IN EXPERIMENTAL NEUROSCIENCE: THE UNRELIABILITY OF FIRING RATE.  (Regarded as a universal truth,  spike timing does not need to be disputed or doubted in neuroscience!)

7.       Aug 20: Dorian Aur: the logical inconsistency of "concept cell " idea is fully revealed based on experimental data.

Indeed the entire group deserves the Ig Nobel Prize  in neuroscience for "breakthrough findings"of "concept cells" discovery. In the last 6-7 years these claims regarding "concept cells" were advertised everywhere by Koch's group. From a unreliable measure (firing rate) they have gone too far. All experiments (Aur et al submitted) show a more powerful model of computation within neurons and in the brain.  The current data analysis of temporal patterns in neuroscience is an insult to the capabilities of actual  neurons to process information. They have completely misled us with temporal coding models regarding semantics or concept cells.  Computation in the brain deserves a specific model, and NED represents is such a model that shapes a new vision in the field. The entire connectionist theory can be seen as a particular model of computing by interaction. I'm planning to write more regarding this relationship. An entire infrastructure needs to be built to allow a different approach.  Neuroelectrodynamics describe computation as an ongoing process shaped by the dynamics and interactions of electric charges. The process of interaction can be evidenced during action potentials and synaptic spikes since transient electrical patterns occur in each generated spike. Intracellularly the coding of information is related to physical machinery able to alter the dynamics of electric charges and their spatial distribution at the molecular level. These subtle changes in single spikes can provide the required information regarding semantics if spike directivity is computed. Therefore, helping each other to see the reality in science is the only thing that makes sense. An idea or hypothesis that is based on errors in methodology or reasoning disappoints sooner or later everyone, everywhere.

The debate is posted  at: http://neuroelectrodynamics.blogspot.com/p/concept-cells.html

Dorian Aur

_______________________________________________
Comp-neuro mailing list
Comp-neuro@neuroinf.org
http://www.neuroinf.org/mailman/listinfo/comp-neuro