Comments on: Baysian perceptions http://charbonniers.org/2009/06/10/baysian-perceptions/ A blog on consciousness by Janet Kwasniak Sat, 31 Jul 2010 06:36:40 +0000 http://wordpress.org/?v=2.2.1 By: daedalus2u http://charbonniers.org/2009/06/10/baysian-perceptions/#comment-644 daedalus2u Wed, 08 Jul 2009 00:08:09 +0000 http://charbonniers.org/2009/06/10/baysian-perceptions/#comment-644 <p><p>But we know that the processing that the brain does is non-algorithmic. .</p><br /> <p>http://www.ncbi.nlm.nih.gov/pubmed/16453069</p><br /> <p>in the sense that the processing can be reversed whereas algorithms (in general) cannot be reversed (e.g. squaring and square root). </p><br /> <p>Brain processing may simulate Bayesian-type computations, but because they are fundamentally non-algorithmic they sort of can't really be Bayesian.</p><br /> JanetK: I tend to agree daedalus2u. I certainly do not think that the brain is formulating equations and solving them to form perceptions. But I do think that the general<br /> ideas of the Baysian approach is very similar to what is happening during perception. We do seem to be forecasting the immediate future and correcting our<br /> perceptions on the basis of mis-match between the prediction and the result.</p> <p>But we know that the processing that the brain does is non-algorithmic. .</p>
<p>http://www.ncbi.nlm.nih.gov/pubmed/16453069</p>
<p>in the sense that the processing can be reversed whereas algorithms (in general) cannot be reversed (e.g. squaring and square root). </p>
<p>Brain processing may simulate Bayesian-type computations, but because they are fundamentally non-algorithmic they sort of can’t really be Bayesian.</p>
JanetK: I tend to agree daedalus2u. I certainly do not think that the brain is formulating equations and solving them to form perceptions. But I do think that the general
ideas of the Baysian approach is very similar to what is happening during perception. We do seem to be forecasting the immediate future and correcting our
perceptions on the basis of mis-match between the prediction and the result.

]]> By: mariana http://charbonniers.org/2009/06/10/baysian-perceptions/#comment-611 mariana Wed, 24 Jun 2009 02:33:40 +0000 http://charbonniers.org/2009/06/10/baysian-perceptions/#comment-611 Sorry janet I tend to respond so late cause this blog does not have a way (at least that I know) to let me know someone responded to the comment. Just wanted to tell you that I just hate the Bayesian hype. Dough its true, many advantages, Quick algorithm, easy to understand, works very well in several cases, for example for nlp, even dough it is not understood why it works well compared to other algorithms that seem more reasonable to be used for that kind of tasks. Besides the point that bayes assumes the independence of the variables, but in real life its used with dependent variables most of the time. Sorry janet I tend to respond so late cause this blog does not have a way (at least that I know) to let me know someone responded to the comment. Just wanted to tell you that I just hate the Bayesian hype.
Dough its true, many advantages, Quick algorithm, easy to understand, works very well in several cases, for example for nlp, even dough it is not understood why it works well compared to other algorithms that seem more reasonable to be used for that kind of tasks. Besides the point that bayes assumes the independence of the variables, but in real life its used with dependent variables most of the time.

]]> By: mariana http://charbonniers.org/2009/06/10/baysian-perceptions/#comment-559 mariana Thu, 11 Jun 2009 08:33:27 +0000 http://charbonniers.org/2009/06/10/baysian-perceptions/#comment-559 <p>I loved the main idea, but every learning algorithm works like that, not only bayesian (and if it is the naive version it is even departing from the wrong assumption that is of the independence of the elements that are being percepted) neural networks, decision trees, support vector machines, etc ...<br /> So why bayesian?</p> JanetK: Mariana, the next post is also on Firth's ideas and is less Bayesian. I am of two minds on this: like the basic Bayesian notion but (like you maybe) I am wary of all the cultish hype around the subject. I loved the main idea, but every learning algorithm works like that, not only bayesian (and if it is the naive version it is even departing from the wrong assumption that is of the independence of the elements that are being percepted) neural networks, decision trees, support vector machines, etc …
So why bayesian?

JanetK: Mariana, the next post is also on Firth’s ideas and is less Bayesian. I am of two minds on this: like the basic Bayesian notion but (like you maybe) I am wary of all the cultish hype around the subject.

]]>