'not sure what you mean by the invariance of consciousness. However, take a look at the vector math in Chris Eliasmith's How to Build a Brain for the construction and comparison of representations.

'not sure what you mean by the invariance of consciousness. However, take a look at the vector math in Chris Eliasmith's How to Build a Brain for the construction and comparison of representations.

I meant the generalised knowledge and explanations about the invariance.The sources usually use each of the visual invariance, the sound invariance and thought patterns have a major effect to build a conscious and briefly mention the differences. So I have said variance for conscious. I thought if I could have a deeper understanding any of them how it does individually, I'd be able to make a comparison such as between sound and visual invariances and think critically to have a perspective on what really matters in the invariance. However, I will read "How to Build a Brain" by Chris Eliasmith. I have seen his speech at TEDx Waterloo and it's interesting, though. Thanks for the recommendation!

Echoing John, am not sure what you mean by 'invariance' - invariance has a very precise definition and meaning in mathematics, and I'm not sure how this meaning carries over into your field.

This being said - going back to the original wording of your question - the brain does not 'discover similarity (or similarities) in the midst of bigdata - What the brain does is seek patterns, and oftentimes it even 'discovers' patterns where there are actually none whatsoever. Hence, the brain is notoriously unreliable in terms of discovering patterns amongst vast amounts of data. 

I am not fond of invariance either, since I am convinced that variability or "just sufficient difference" is the motor of perception. What lies behind this perhaps confusing expression is that the brain is always busy to propose - thereby anticipate - what "it is likely to be". A matching with ongoing sensory data - like in the efference copy  - provides in the same time an indication of how relevant this anticipation is.

In this way, I would say that the brain is always in need for variance or discrepancy. This is the information quest, as illustrated by curiosity: we react to the unknow in reducing uncertainty, but we search for the unknown.

Hope my position is clear enough. If not and you nevertheless wish to pursue, I will pursue.

PS: to Chris, I would say that creating a pattern that does not exist corresponds to what is called a working hypothesis in need of confirmation. It fits "to a certain extent" to possible patterns. Is it creativity?

I would say that at the neuronal level, there is an effect I call similarity selection that matches similar stimuli. I am not sure it is Invariance either, since it happens in an uncertain milleu where each neuron is constantly morphing.

I think this is an interesting question and I think the answer to the 'subtitle' question is that there are no good sources on this. The neurocomputational people have simply not addressed the problem. I think I have to disagree with Chris and agree with Françoise and Graeme that the brain must discover similarity but I have never seen a proposed mechanism. People talk of efferent copies with the assumption that the brain generates a mock up of what it expects to receive in the way of patterns in the context of some ocular or bodily movement and then compares the mock up pattern with actual input. But nobody much says how the comparison is done.

My own thought is that it requires a form of integration that subtracts. You cannot compare pattern 3  6  4  8  with pattern 3  6  4  8 or perhaps more interestingly 3x1  3x2 2x2 4x2 by summation. But you can see they are the same by subtraction. The only place where computations actually occur in brains is in the dendritic trees of individual neurons - post-synaptic integration. So it seems we need a subtractive integration. That could be done by matching inhibitory inputs with excitatory inputs but the evidence so far suggests that inhibitory and excitatory inputs serve different function - inhibitory ones being mostly feedback signals to prevent overexcitation and mixing of independent data streams. So I strongly suspect that subtraction is done by some sort of 'balancing' of paired excitatory inputs coupled to a causal step that has not yet been considered by the post synaptic integration gurus. Recent work on the importance of temporal offsets in integration might be a clue.

The other point in the question, which I think is relevant, is how the brain ensures that it 'cleans up' the pattern it wants to compare in the sense of cutting out everything it does not want to be salient. I suspect that involves triage by synchrony, so we would expect the brain to ensure that the two patterns it wants to compare are fed into a dendritic tree (or each of many) synchronised both within themselves and to the partner.

I think invariance will come in later. Cells that receive patterns with the job of identifying or 'labelling' those patterns can have outputs that simply indicate the label. So if my brain expects an en face rose shape at the fovea having looked left and it finds an en face rose shape at the fovea some cells will confirm that this is a rose and send a 'rose' message. Other cells in other situations might be set up to confirm a profile rose shape. If the outputs connect to cells that carry the memory reference tag for 'rose invariant' i.e. roseness in whatever position their outputs can feed through to cells involved in more abstract thoughts and memories of which seasons roses appear or what they smell like etc. 

Unfortunately almost none of this seems to get discussed in specific computational terms in neurobiology papers because nobody wants to attribute meaning to individual cells. I think it is time we did again. Even Horace Barlow, who steered people away from meaning in individual cells in 1972 has changed his mind as I understand it.

In my theory, subtractive technology is built into the neuron. While it seems obvious that the integration is additive, using a sort of wired or technology, there is actually the capability to "Demorganize" a neuron using negation. A generally hyperactive neuron, with tonic inhibition and an inhibited input will create a NAND like effect.

This is a negative logic selective neuron.

Thank you so much, everyone, it's just much clearer in my mind. Jonathan Edwards, it was exactly the point I am asking about. I have seen that there are different approaches rather than an accepted unifying framework.