Part 5 : Connectome

5.1. Neural connections

A Connectome is the map of the connections between neurons in the brain.

These inter-neuronal connections facilitate in encoding stimuli, deciphering stimuli and shaping the lifeform's overall behaviour. All incoming stimuli is encoded in the brain in the form of connections between neurons.

When a new memory is formed, it is done so through modification of the weights of the inter-neuronal connections. Impulses to initiate and implement specific actions are encoded by these connections.

5.2. Two Types of Connections

In this model, Connections are differentiated into two categories :

1. hardwired connections , which are stable and don't strengthen/weaken

2. dynamic connections , which are created based on neural activity and allowed to strenghten & weaken.

5.3. Innate (hardwired) connections for survival

Neuronal connections which enable food-seeking and danger avoidance aid the survival of the lifeform. When a biological organism has just been born, it has to survive in a specific environment, and innate connections related to food-seeking/danger-avoidance behaviour aid in its survival.

Such innate connections enable "Primitive reflexes" ^[2] , which is the execution of certain motor action sequences on encountering a particular stimuli. For example, in human infants, rooting reflex, stepping reflex, palmar grasp reflex,etc. are observed. After sometime after birth, such reflexes either come under voluntary control, or fade away.

Therefore any Artificial Lifeform should be preloaded with some innate connections to enable survival in the specific environment it is supposed to be 'placed'.

5.4. Creation of new dynamic connections

Sensory organs receive the stimuli many times per second and transmit the signals to the brain simultaneously as a continuous datastream. The brain receives sensory datastreams of various types (vision,sound,etc.) simultaneously. Regions of the brain are specifically mapped to sensory type (eg: visual stimuli mapped to occipital lobe). To form an association between stimuli, connections need to be formed between the corresponding neurons that encode the different stimuli.

Let H be the maximum delay allowed in forming connections ie., connections with a delay of upto t+H synaptic timesteps are allowed in the brain.

Criteria for new connection formation :

In this model,criteria for forming new connections will be based on Hebbian learning, which can be loosely stated as "Neurons that fire (almost) together, wire together".
If neuron A and neuron B are activated within a short time (within H synaptic timesteps of each other), then a dynamic connection is created between neuron A and neuron B. (provided that neural pathway is permitted)
For example, if neuron A fires at time t and neuron B fires at t+X synaptic timestep, a (t+X) dynamic connection is created from neuron A to neuron B. (where 1<=X<=H )

5.5. Strengthening of existing dynamic connections

Criteria for strenghtening an existing dynamic connection :

When neuron A fires at time t , and neuron B fires at timestep t+X , and a dynamic t+X connection already exists, then the connection strength of the t+X connection from A to B will be increased at each instance, upto a set limit. (where 1<=X<=H)

5.6. Weakening of existing dynamic connections

Criteria for weakening an existing dynamic connection :

Hardwired connections are deemed stable and not strengthened or weakened.
Whenever a neuron fires, all its incoming dynamic connections are weakened except those dynamic connections which caused it to fire. This criteria is based on known research findings^[r3]

5.7. Explaining Classical conditioning using this model

Using the above criteria for dynamic connection creation/strenghtening/weakening, it is possible to explain how (Pavolovian) classical conditioning can occur.

In the following two examples,
• Let neuron "A" be the "Unconditioned Stimulus" and "neuron C" firing be the "Unconditioned Response".
• Let neuron B be the "neutral stimulus"
• When the Unconditioned stimulus A and neutral stimulus B are fired simultaneously, a (t+X) connection forms between the neutral stimulus B and the unconditioned response C.

Interactive examples

Classical conditioning : Example 1

In this example, H=1, so upto (t+1) connections are allowed to form.

Step 1: Click "Activate both A and B simultaneously": This creates a new connection from B to C , and the connection weight w strenghtens on each instance, upto w=3.

Step 2: Click "Activate only B" : Using the newly formed connection, triggering B alone will cause C to fire.

Step 3: Click "Activate only A" : Since A causes C to fire at time t without B being active at (t-1), the criteria for weakening is satisfied, and therefore the strength of the t+1 dynamic connection from B to C is weakened repeatedly, until it is removed.

Classical conditioning : Example 2

In this example, H=2, so upto (t+2) connections are allowed to form.
An additional neutral stimulus B1 is introduced, and is excited simultaneously as the Unconditioned stimulus A and netural stimulus B.

Step 1: Click "Activate A, B and B1 simultaneously": This creates new t+1 &t+2 connections from B to C and B1 to C, and the connection weight w strenghtens on each instance, upto w=3.

Step 2: Click "Activate only B" : Using the newly formed connection, triggering B alone will cause C to fire. At the same time, since B1 didn't cause C to fire, the dynamic connection from B1 to C is weakened repeatedly, until it is removed.

Step 3: Click "Activate only A" : Since A causes C to fire at time t without B being active at (t-1) & (t-2) , the criteria for weakening is satisfied, and therefore the strength of the t+1 and t+2 dynamic connection from B to C is weakened repeatedly, until it is removed.