Scientific Understanding of Consciousness
Consciousness as an Emergent Property of Thalamocortical Activity

Autoassociation Diagram

(Rolls & Treves; Neural Networks and Brain Function, 42)





Autoassociation Memory

Autoassociative memories, or attractor neural networks, store memories, each one of which is represented by a pattern of neural activity. They can then recall the appropriate memory from the network when provided with a fragment of one of the memories. This is called completion. Many different memories can be stored in the network and retrieved correctly. The network can learn each memory in one trial. Because of its 'one-shot' rapid learning, and ability to complete, this type of network is well suited for episodic memory storage, in which each past episode must be stored and recalled later from a fragment, and kept separate from other episodic memories. An autoassociation memory can also be used as a short term memory, in which iterative processing round the recurrent collateral connection loop keeps a representation active until another input cue is received. In this short term memory role, it appears to be used in the temporal visual cortical areas with their connections to the ventrolateral prefrontal cortex for the short term memory of visual stimuli; and in the dorsolateral prefrontal cortex for short term memory of spatial responses. A feature of this type of memory is that it is content addressable; i.e., the information in the memory can be accessed if just the contents of the memory (or a part of the contents of the memory) are used. This is in contrast to a conventional computer, in which the address of what is to be accessed must be supplied, and used to access the contents of the memory. Content addressability is an important simplifying feature of this type of memory, which makes it suitable for use in biological systems.

Architecture and operation

The prototypical architecture of an autoassociation memory is shown in the diagram. The external input ei is applied to each neuron i by unmodifiable synapses. This produces firing ri of each neuron, or a vector of firing on the output neurons r. Each output neuron i is connected by a recurrent collateral connection to the other neurons in the network, via modifiable connection weights wy. This architecture effectively enables the output firing vector r to be associated during learning with itself. Later on, during recall, presentation of part of the external input will force some of the output neurons to fire, but through the recurrent collateral axons and the modified synapses, other neurons in r can be brought into activity. This process can be repeated a number of times, and recall of a complete pattern may be perfect. Effectively, a pattern can be recalled or recognized because of associations formed between its parts. This of course requires distributed representations.


Return to Memory

CA3 subregion of the hippocampus. (Rolls; Memory, Attention, and Decision-Making, 94)

CA3 as an Autoassociation Memory (Rolls; Memory, Attention, and Decision-Making, 58)

A theory of the hippocampus in which the CA3 neurons operate as an autoassociation memory to store episodic memories including object and place memories, and the dentate granule cells operate as a preprocessing stage by performing pattern separation so that the mossy fibers could act to set up different representations for each memory to be stored in the CA3 cells. (Rolls; Memory, Attention, and Decision-Making, 37)

The hypothesis is that the autoassociation memory enables arbitrary sets of concurrent neuronal firings, representing for example the spatial context where the episode occurred, the people present during the episode, and what was seen during the episode, to be associated together and stored as one event. (Rolls; Memory, Attention, and Decision-Making, 41)

The final convergence afforded by the hippocampus into one network in CA3 may be especially appropriate for episodic memory typically involving arbitrary associations between the any of the inputs of the hippocampus, e.g. spatial, vestibular related to self motion, visual object, olfactory, and auditory. (Rolls; Memory, Attention, and Decision-Making, 39)

A fundamental property of the autoassociation model of the CA3 recurrent collateral network is that it can implement a short-term memory by maintaining the firing of neurons using the excitatory recurrent collateral connections. (Rolls; Memory, Attention, and Decision-Making, 62)

The three major classes of excitatory input to the CA3 cells (recurrent collateral, mossy fiber, and perforant path) could be independently scaled by virtue of the different classes of inhibitory interneurons. (Rolls; Memory, Attention, and Decision-Making, 74)

The dentate granule cell stage of hippocampal processing, which precedes the CA3 stage, acts to produce during learning the sparse yet efficient representation in CA3 neurons that is required for the autoassociation to perform well. (Rolls; Memory, Attention, and Decision-Making, 75)