Artificial Consciousness/Neural Correlates/Neural Network Models/Learning new connections
Learning New Connections[edit | edit source]
Although a fully connected network is the most compact network, There might be a case for redundant connections, especially if the connections are Stochiatically organized. From the point of view of an organism that has limited resources it seems reasonable to double up on important connections and prune less important connections. This has in the past been called a Branch and Prune strategy.
Unfortunately determining the best strategy of branch and prune has proven to be difficult. It is only recently that we have begun to recognize the Molecular Traces associated with growing new neural fibrils, and/or absorbing them. The bio-molecular traces are complex, and operate in parallel in a chemical cascade reaction that we do not fully understand yet, so this model must be seen to be speculative, and therefore will probably need to be replaced once the science is further along.
Currently what we think is happening is that the chemical cascade reaction is sensitive to the activity of the synapses, and therefore preferentially results in new growth local to active synapses. Theoretically this can be either the growth of new synapses, or the growth of new fibrils to connect the neurons. We don't yet know the full chemical story, so we don't know yet whether we will find evidence of steering chemicals that guide the fibrils towards new active synapses, but we can assume from the sheer indiviuality of the connections that connections are opportunistic and are not specified as to end location, and thus can only be guided stochiastically. The brain goes through different growth rates, which we think are mediated partially by the presence of Growth Factors, which mean that the prevalence of opportunistic connections peaks at a relatively young age, and that the adult brain has a considerably smaller number of connections than the childs brain. What we think this means is that the brain does both branching and pruning and that pruning overtakes branching at a certain age.
Perhaps the reason that the brain has fewer connections in the adult brain is that the brain is more finely focussed on what it remembers, or perhaps adults just tend to reduce the variety of the things they learn compared to children. In any case Plasticity or the ability to make new connections tends to gutter out at about 60 years old, with the loss of Growth Factors that are related to cell fibril proliferation. Recent experiments with hormone replacement have shown that the age of senescence can be pushed back as long as hormones are supplied.
It is thought that the sensitivity to activation of synapses, might be related to the sequestering of protein tags associated with the synapse, if so, we will probably find that the chemical cascade reactions preferentially extend where there is a protein tag of a certain type.