Semblance hypothesis was developed based on the assumption that memory is a virtual sensation of certain sensory inputs. This makes it difficult to understand the nature of the retrieved memories by reductive approaches using biological experiments. Computational studies, though challenging, may offer alternate ways to understand and interpret semblances that produce memory. Simultaneous fine tuning of a very large number of variables, described in the hypothesis, will be necessary for understanding the nature of semblances and this may be achieved by computational modeling.

For details of approaching the problem, please refer to the book here. A step wise approach that may solve the problem may be carried out by using the following steps.

1.   Assign different sensory receptors and neuronal orders. Postsynapses of the synapses at the fifth to seventh neuronal orders from different sensations can be allowed to converge at the hippocampus.

2.  Based on the closeness of the postsynapses and their simultaneous activation, functional LINKs are allowed to form during learning.

3.  Based on the assignment of the positions of the postsynapses, we can assign values of semblances at each postsynapse.

4.  A rigid model is needed to be created initially before adding the flexibility due to synapse elimination, synapse addition and hippocampal new neuron formation.

5.  Given the fact that the structure of the artificial nervous system is known, make a list of all the potential functional LINKs that can be formed provided there is simultaneous activation of the postsynapses that can LINK them during various associative learning sessions.

6.  Make a list of all the network semblances that can be formed.

7.  Identify the synaptic semblance values that can be formed at each of the LINKable postsynapses.

8.  Identify the potential network semblance values for each neuron based on the LINKable positions of their dendritic spines (postsynapses). This can provide the nature of the network semblances.

9.  Learning is carried out by using a cue and an item to be learned. Expected functional LINKs that are likely to be formed between the postsynapses of the neurons belonging to the cue and the item at different neuronal orders (where the synapses belonging to the cue and the learned item come to functionally LINKable distances) are marked.

10. Since learning in an adult nervous system depends on all the structural and functional LINKs already existing in the brain, we may need to induce several associative learning (and therefore functional LINKs) before we can expect efficient memory retrieval using a new learning test that we are planning to execute.

11.  We need to standardize a large number of variables for a given nervous system in order to understand the mechanism completely. These include a) required prior associative learning and the number of functional LINKs formed from them b) nature of the theta/gamma rhythms that can provide 1) reactivation of already formed functional LINKs 2) sub-threshold activation of many neurons at higher orders 3) new neuron formation and change in semblance locations that occur during transfer of memories from the hippocampus to the cortex (see section III.9).

12.  Once sufficient learning is done, introduce a specific cue for retrieving the memory. Identify all the specific synaptic and network semblances that occur at the LINKed postsynapses in the presence of the cue.

13.  Plot the sensory neuron location map from each of the synaptic and network semblances. Next, integrate these sensory neuron location maps and make a sensory identity map of the item retrieved (Figs.9, 10). Extrapolate the sensory identity of the item retrieved. By comparing its characteristic features with that of the learned item, fine-tune the artificial nervous system to obtain a good match of the sensory density map. This fine-tuning needs to be carried by changing the factors given in Table 1. These factors can lead to certain level of fluctuation of the identity of the retrieved memory using a given cue over time due to addition and deletion of synapses, addition of new neurons, and formation of new functional LINKs from new learning events.

Since there are many factors that need to be fine-tuned at a given time, it will require a huge computational effort. A novice artificial nervous system (corresponding to a nervous system belonging to a child in the first few years of life) designed using computational models will have (S→F) LINKs (see section II.8.3 in the book) to maintain the basic brain functions. We need to teach the artificial nervous system by exposing it to simultaneous sensory stimuli for associative learning prior to obtaining any useful output during memory retrieval. This is primarily due to the fact that each associative learning uses many functional LINKs formed during prior associative learning. Many of the learning can utilize transferability property of the functional LINKs at different areas of the brain which is a function of the architecture of a given nervous system.

Different cue stimuli will induce the same synaptic semblance when they pass through one LINKed postsynapse to the postsynapse that belongs to the item to be retrieved. One cue stimulus activates a synapse in an islet of postsynapses and spreads through the functional LINKs to produce different synaptic semblances. Many of these activated postsynapses within an islet of functional LINKs do not belong to/represent the learned item. The synaptic semblances resulting from their activation will have only negligible contribution towards the net semblance for the memory to be retrieved.

Since the number of LINKable postsynapses is finite and the item to be memorized can be complex, learning capacity will be determined by the number of specific combinations/ permutations of synaptic and network semblances. Learning of related items will require newer items for associative learning through the functional LINKs in addition to activations of previously induced functional LINKs to reach the threshold semblance for their retrieval. Newly formed neurons can bring an additional capacity for combinations or permutations of semblances at the orders of neurons above their level and can create new neuronal network semblance providing the specificity required for memory for additional items to be memorized using the same sensory system. This depends on the percentage contribution of transmission of activities through the new neurons.

Computational approaches may provide feasibility to examine the effect of multiple variables at a given time to optimize the required semblance output. Eventually we hope that the computational fine-tuning will be able to provide answers to the following questions.

1.   What is the required number of the functional LINKs at each neuronal order for effective semblance?

2.  What is the extent of the partial neuronal networks for effective semblance?

3.  What weights are required to be given for synaptic and network semblance for effective memory?

4.  At what neuronal orders should the axons converge for potential functional LINK formation during learning in a given nervous system?