1 SCIENTIFIC REPORTS | (2018) 8:1403 | DOI:10.1038/s41598-018-19462-3 www.nature.com/scientificreports Random neuronal ensembles can inherently do context dependent coarse conjunctive encoding of input stimulus without any specifc training Jude Baby George 1 , Grace Mathew Abraham 1 , Zubin Rashid 1 , Bharadwaj Amrutur 2 & Sujit Kumar Sikdar 3 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 frst 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 classifcation without needing any specifc 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 sufcient connectivity creates similar coarse-conjunctive encoding of input sequences. Pattern or sequence recognition and classifcation is a well-studied problem in engineering that uses biologically inspired architectures like artifcial neural networks, and more recently deep learning networks that have shown promising results in solving such tasks. However, the learning algorithms adopted by these architectures require multiple iterations and modifcations of the connectivity weights across all layers of the network. Te existence of similar multi-layered learning in the biological neuronal networks for efcient processing of input stimuli and classifcation of inputs has not been observed yet experimentally. An alternative learning architecture is to have a random neuronal ensemble with a mix of inhibitory and excitatory neurons that is then connected to another layer of perceptron type neurons, in a probabilistic manner, with learning restricted to the fnal perceptron layer. We describe this further in the schematic in Fig. 1, where a layered neuronal system with probabilistic connec- tivity at input and output of frst layer, is connected to a second layer having neurons equipped with STDP, to solve the problem of input classifcation without any need for network modifcation/learning at the input layer. We experimentally validate this architecture by using neuronal ensembles cultured on a multi electrode array, to form the frst layer of the Fig. 1. Te multi-electrode array allows us to create complex spatio-temporal input stimulation patterns, that get encoded by the neuronal tissue which is then observed as responses at the electrodes for further analysis. We show through modeling and by ftting experimental data that probabilistic connections and a layered architecture as in Fig. 1, can provide a very robust platform to implement context dependent clas- sifcation. Our data and results show the presence and usefulness of coarse-conjunctive tuning of neurons in the 1 Center for Nanosicence and Engineering, IISc Bangalore, Bengaluru, Karnataka, India. 2 Robert Bosch Center for Cyber-Physical Systems and Department of Electrical Communications Engineering, IISc Bangalore, Bengaluru, Karnataka, India. 3 Molecular Biophysics Unit, IISc Bangalore, Bengaluru, Karnataka, India. Correspondence and requests for materials should be addressed to S.K.S. (email: sks@iisc.ac.in) Received: 2 May 2017 Accepted: 14 December 2017 Published: xx xx xxxx OPEN