Here is our commentary on the paper
The Schurger et al. response
Tuesday, May 25, 2010
Subscribe to:
Post Comments (Atom)
skip to main |
skip to sidebar
This blog is meant to be a forum for discussion about issues involved in using MVPA (multivariate pattern analysis) and PD (predictive decoding) techniques.
The Schurger Response makes two arguments to deflect my concerns:
ReplyDelete(1) Doing Cross validation protects you from BIAS.
This is simply false. If you have a biased measure as I attempted to argue in my commentary, you cannot protect yourself by taking more SAMPLES of the BIAS. You will simply reduce the standard error on the sampling distribution of those BIASED sample estimates. You will still retain the BIAS.
The only way to reduce BIAS is to make sure you have sufficient model selection cases which may be more appropriate for your data. It is too bad that they didn't try some simple nonlinear bias in theoriginal classification that may have produced astatistically significant result in the first place and then avoided the apparent ad hoc nature of there derived variances measures.
(2) That they didn't really need to use a classifier as they got the same results with the GLM. But this is particularly odd, since they
predicated the result they found with the purported sensitivity of the classifier and their new measure of voxel variance: "angular deviation" which was selected by taking significant folds from the classifier! SO how then do you get the
same exact result without the classifier? What was the point of using it in the first place?
Their claim in their response to my commentary was: "to determine the number of voxels to keep for training"... which in turn determined their
crossvalidation results, which in turn determined their measure of angular deviation. Consequently,
propagating the BIAS throughout their analysis.
Again, this paper would appear to be a classic case of how to (mis)-use MVPA and classifier methods. It is critical to be able at this early date to separate the wheat from the chaff, else the field itself will propagate silly biases.
QBT: Nano-synaptic - communication & creativity
ReplyDeleteThe traditional model of communication synapse is too slow, and localized and non-interactive to still be useful in interpreting the intuition and creativity;
On September 4, 2014 14:13, Paul Manz wrote:
At the base of the project Quantum Creativity is the idea of proposing a model QBT (Quantum Brain Theory) in the synapses of the brain activity as it is necessary to explain the simultaneous communication in a synergistic way , since that it is the property of creative thinking. This idea may be developed through the extension of quantum science to the functioning of the brain and in particular through the introduction of entanglement as a possible solution to improve understanding of the plasticity of the brain, which is the basis of creativity '.
In fact, currently the functionality of neurotransmitter released in the synaptic cleft it is only interpreted so as to implement a temporary bond with postsynaptic receptors. The change of potential caused by the displacement of ions CA ++, allows the transfer of an information signal. Subsequently to allow a new variation of the local synaptic potential and transfer of a pulse of information it becomes a need to renewed electro-ion potential change so as to allow a subsequent transfer of one way information, between two neurons Therefore it becomes necessary that the temporary bond of the neurotransmitter should be broken and the neurotransmitters in the synaptic cleft necessarily have to be degraded by enzymes or resorbed, in order that the electron-ion potential can notbe stabilized, because 'what this block any other possible variation of the of information pulses.
The traditional model of the activity of synaptic neurotransmission, here shortly summaryzed , it is well known, and it is largely theoretical, since the size of the nano-synaptic gaps, does not allow a precise visibility' of its dynamics. In addition this traditional model of the communication by means of chemical synapses has the serious limitation of being very slow in relation to the activity of thought. In essence this traditional model which is based on a local and unidiretional paths, it is not interactive, consequently does not allow to conceive the synergistic actions delocalized over vast areas of the brain which act simultaneously and synergistically, as would be desirable to obtain a better understanding of creativity of thought.
Till today the traditional neuro-science still seems to refer to such a simple but rather complicated model of an unidirectional& mechanical explanation of the activity of neurotransmitters in the chemical synapses. Such an explanation therefore leads to misinterpretation on the way of evoluttion of functioning of the brain, as they are based on a communication system that passively reconstructs information from uni-directionally to 'environment. The above leads to conceive cognitive extrapolations very questionable such as the objectivity of perception, which ultimately tends to allow the genesis of false ideas about the thought and his creative development.
An innovative model QBT "based on quantum science" by introducing the concept of quantum entanglement in 'communication activities of nano-synapses, will give an added value to the understanding of various puzzles (such as optical illusions, the visual perception during the dream. ..ecc ...) which to date are not easily interpretable by following the simplistic classical model but for the project quantum creativity would be more important to understand as fundamentals fof quantum cognition : intuition, creativity and holistic consciousness.
Finally I would like to know what partners intend to do to create a highly collaborative model of Quantum Brain Theory (QBT), which may be related with the classical model in order to highlight the limitations but also the possible complementarity during the development of the project on Quantum Creativity.