@STRING{colt93 = "Proceedings of the Sixth Annual ACM Conference on Computational Learning Theory" } @STRING{jacm = "Journal of the Association for Computing Machinery" } @STRING{jma = "Journal of Multivariate Analysis" } @STRING{lncs = "Lecture Notes of Computer Science, Springer" } @STRING{lncs = "Lecture Notes in Computer Science" } @STRING{nips = "Advances in Neural Information Processing Systems" } @STRING{pecs = "Colloquia Mathematica Societatis Janos Bolai, 57.\ Limit Theorem in Probability and Statistics, Pecs (Hungary)" } @STRING{spl = "Statistics \& Probability Letters" } @MastersThesis{Joshi:02, author = {P. Joshi}, title = {Synthesis of a Liquid State Machine with Hopfield/Brody Transient Synchrony}, school = {Center for Advanced Computer Studies, University of Louisiana, Lafayette, USA}, year = 2002, month = {November}, abstract = {Understanding the mechanism of spatiotemporal integration used by our brain to perform recognition of complex temporal sequences is a challenge for current researchers in neuroscience. Recent research has proposed transient synchrony as a plausible mechanism for spatiotemporal integration. This thesis studies a biologically plausible network architecture made of simulated minicolumns that performs a temporal integration task, specifically spoken-word recognition. The network's ability to recognize a spoken word and its natural variants is independent of variations across speakers, simple masking noises and variations in system parameters. The network demonstrates inter columnar and intra columnar synchrony, which in turn leads to word recognition. The intra columnar synchrony of minicolumns acts as an event detection mechanism for events in a particular frequency band. The inter columnar transient synchrony enables the network to recognize words. Each of the minicolumns exhibit a very unique temporal signature when presented with a temporal input. These signatures looked nearly the same for similar inputs (e.g., the same word spoken by different speakers etc.) and were strikingly different for different temporal inputs (e.g., different words).} } @InProceedings{Joshi:06, author = {P. Joshi}, title = {Modeling working memory and decision making using generic neural microcircuits}, editor = {Stefanos Kollias and Andreas Stafylopatis and W{\l}odzis{\l}aw Duch and Erkki Oja}, booktitle = {Artificial Neural Networks -- ICANN 2006}, pages = {515--524}, year = {2006}, volume = {4131}, series = {Lecture Notes in Computer Science}, publisher = {Springer}, isbn = {3-540-38625-4}, abstract = {Classical behavioral experiments to study working memory typically involve three phases. First the subject receives a stimulus, then holds it in the working memory, and finally makes a decision by comparing it with another stimulus. A neurocomputational model using generic neural microcircuits with feedback is presented here that integrates the three computational stages into a single unified framework. The architecture is tested using the two-interval discrimination and delayed-match-to-sample experimental paradigms as benchmarks.} } @Article{Joshi:06b, author = {P. Joshi}, title = {From memory based decisions to decision based movements: A model of interval discrimination followed by action selection}, journal = {Neural Networks}, year = {2007}, publisher = {}, note = {in press}, abstract = {Interval discrimination task is a classical experimental paradigm that is employed to study working memory and decision making and typically involves four phases. First the subject receives a stimulus, then holds it in the working memory, then makes a decision by comparing it with another stimulus, and finally acts on this decision, usually by pressing one of the two buttons corresponding to the binary decision. A neurocomputational algorithm using generic neural microcircuits with feedback is presented here that integrates the four computational stages into a single unified framework. The algorithm is tested using two-interval discrimination and delayed-match-to-sample experimental paradigms as benchmarks.} } @PhDThesis{Joshi:07, author = {P. Joshi}, title = {On the role of feedback in enhancing the computational power of generic neural microcircuits}, school = {Graz University of Technology}, year = 2007, abstract = {Circuits of neurons in the brain perform diverse cortical computations in parallel, endowing the organism with diverse cortical modalities, e.g. motor control, vision, and audition; and higher order cognitive processes, e.g. planning, and decision making. It is believed that these computations are carried out by network of neurons in cortical microcircuits, where each microcircuit is composed of rather stereotypical circuit of neurons within a cortical column. A characteristic property of these cortical circuits is the presence of abundant feedback connections, be it on the level of recurrent axon collaterals projecting back onto the same neuron, or on the network level between different cortical areas. This thesis explores the functional role of neural feedback in enhancing the computational power of generic neural microcircuits. It is shown that feedback endows standard models for neural circuits with the capability to emulate arbitrary Turing machines. In fact, with a suitable feedback such circuits can simulate any dynamical system, in particular any conceivable analog computer. Under realistic noise conditions the computational power of these circuits is obviously reduced. However it is demonstrated through computer simulations that feedback also provides a significant gain in computational power for quite detailed models of cortical microcircuits with in-vivo-like high levels of noise. Furthermore neurocomputational models using generic neural microcircuits with feedback are explored in the context of motor control, decision making, and ``action selection in presence of decisions''.} } @InProceedings{JoshiMaass:03, author = {P. Joshi and W. Maass}, title = {Movement Generation and Control with Generic Neural Microcircuits}, booktitle = {Biologically Inspired Approaches to Advanced Information Technology. First International Workshop, Bio{ADIT} 2004, Lausanne, Switzerland, January 2004, Revised Selected Papers}, pages = {258--273}, year = {2004}, editor = {A. J. Ijspeert and M. Murata and N. Wakamiya}, volume = {3141}, series = {Lecture Notes in Computer Science}, publisher = {Springer Verlag}, abstract = {Simple linear readouts from generic neural microcircuit models consisting of spiking neurons and dynamic synapses can be trained to generate and control basic movements, for example, reaching with an arm to various target points. After suitable training of these readouts on a small number of target points; reaching movements to other target points can also be generated. Sensory or proprioceptive feedback turns out to be essential for such movement control, even if it is noisy and substantially delayed. Such feedback turns out to optimally improve the performance of the neural microcircuit model if it arrives with a biologically realistic delay of 100 to 200 ms. Furthermore, additional feedbacks of ``prediction of sensory variables'' are shown to improve the performance significantly. The proposed model also provides a new approach for movement control in robotics. Existing control methods in robotics that take the particular dynamics of the sensors and actuators into account (``embodiment of robot control'') are taken one step further by this approach, which provides methods for also using the ``embodiment of computation'', i.e. the inherent dynamics and spatial structure of neural circuits, for the design of robot movement controllers.} } @Article{JoshiMaass:04, author = {P. Joshi and W. Maass}, journal = {Neural Computation}, title = {Movement Generation with Circuits of Spiking Neurons}, year = {2005}, volume = 17, number = 8, pages = {1715--1738}, abstract = {How can complex movements that take hundreds of milliseconds be generated by stereotypical neural microcircuits consisting of spiking neurons with a much faster dynamics? We show that linear readouts from generic neural microcircuit models can be trained to generate basic arm movements. Such movement generation is independent of the arm-model used and the type of feedbacks that the circuit receives. We demonstrate this by considering two different models of a two-jointed arm, a standard model from robotics and a standard model from biology, that each generate different kinds of feedback. Feedbacks that arrive with biologically realistic delays of 50--280 ms turn out to give rise to the best performance. If a feedback with such desirable delay is not available, the neural microcircuit model also achieves good performance if it uses internally generated estimates of such feedback. Existing methods for movement generation in robotics that take the particular dynamics of sensors and actuators into account (``embodiment of motor systems'') are taken one step further with this approach, which provides methods for also using the ``embodiment of motion generation circuitry'', i.e., the inherent dynamics and spatial structure of neural circuits, for the generation of movements.} } @InProceedings{MaassETAL:06, author = {W. Maass and P. Joshi and E. D. Sontag}, title = {Principles of real-time computing with feedback applied to cortical microcircuit models}, booktitle = {Advances in Neural Information Processing Systems}, abstract = {The network topology of neurons in the brain exhibits an abundance of feedback connections, but the computational function of these feedback connections is largely unknown. We present a computational theory that characterizes the gain in computational power achieved through feedback in dynamical systems with fading memory. It implies that many such systems acquire through feedback universal computational capabilities for analog computing with a non-fading memory. In particular, we show that feedback enables such systems to process time-varying input streams in diverse ways according to rules that are implemented through internal states of the dynamical system. In contrast to previous attractor-based computational models for neural networks, these flexible internal states are {\em high-dimensional} attractors of the circuit dynamics, that still allow the circuit state to absorb new information from online input streams. In this way one arrives at novel models for working memory, integration of evidence, and reward expectation in cortical circuits. We show that they are applicable to circuits of conductance-based Hodgkin-Huxley (HH) neurons with high levels of noise that reflect experimental data on in-vivo conditions. }, year = {2006}, volume = {18}, pages = {835--842}, publisher = {MIT Press}, editor = {Y. Weiss and B. Sch\"olkopf and J. Platt} } @Article{MaassETAL:06a, author = {W. Maass and P. Joshi and E. D. Sontag}, title = {Computational aspects of feedback in neural circuits}, journal = {PLOS Computational Biology}, year = 2007, volume = {3}, number = {1}, pages = {e165, 1--20} }