Random neuronal ensembles can inherently do context dependent coarse conjunctive encoding of input stimulus without any specific training

Abstract

Conjunctive encoding of inputs has been hypothesized to be a key feature in the computational capabilities of the brain. This has been inferred based on behavioral studies and electrophysiological recording from animals. In this report, we show that random neuronal ensembles grown on multi-electrode array perform a coarse-conjunctive encoding for a sequence of inputs with the first input setting the context. Such an encoding scheme creates similar yet unique population codes at the output of the ensemble, for related input sequences, which can then be decoded via a simple perceptron and hence a single STDP neuron layer. The random neuronal ensembles allow for pattern generalization and novel sequence classification without needing any specific learning or training of the ensemble. Such a representation of the inputs as population codes of neuronal ensemble outputs, has inherent redundancy and is suitable for further decoding via even probabilistic/random connections to subsequent neuronal layers. We reproduce this behavior in a mathematical model to show that a random neuronal network with a mix of excitatory and inhibitory neurons and sufficient connectivity creates similar coarse-conjunctive encoding of input sequences.

Working Memory for Online Memory Binding Tasks: A Hybrid Model

“…We create our simple hybrid model to separate temporary memory (BRN) from the executive part (FFN). Notwithstanding, the biological plausibility of BRN and its capability to hold and transform temporary information is validated in many areas of the brain [6,10,11,14,21,22,24,26]. Our basic binding memory tasks aim to show the meaning of these functional features of our model of WM.…”

“…Although the input shape to BRN is complex, you can see the effect of linear conjunctions in BRN's output. So, BRN encodes inputs to outputs as expected [11,27].…”

“…As you can see the fluctuation in loss function is much larger compared with previous convergence (Figs. 9,11). This could result from BRN sensitivity to inputs and its complex chaotic activity which makes FFN prediction harder to archive [21,26].…”

“…Most of these transformations require learning over random networks to decode temporary evolving representations which make them inflexible for WM [6]. Indeed, random networks can inherently maintain information and encoding of input stimulus without any specific learning [6,11]. There is also evidence for similar functioning of WM in the prefrontal cortex [18].…”

Precise spatial spatial memory in local random networks

“…Yet it is not clear that networks with random weights, or unstructured connectivities, can perform similar computations. Indeed, random excitatory-inhibitory networks have been shown to be capable of various complex computations, including conjunctive encoding for input classification [16] and, in the balanced case, emergent selectivity in the context of evidence integration tasks [17].…”