Arminda Moreno-Díaz

Approximate Solutions to Semi Markov Decision Processes through Markov Chain Montecarlo Methods

We explore the possibilities of Markov Chain Monte Carlo simulation methods to solve sequential decision processes evolving stochastically in time. The application areas of such processes are fairly wide, embedded typically in the Decision Analysis framework, such as preventive maintenance of systems, where we shall find our illustrative examples.

Bioinspired computing nets for directionality in vision

Directional selectivity to local visual stimuli appears in various levels of the visual pathway, being in the retina very conspicuous. Neurophysiology suggests that directionality (as well as other local and quasi-global filtering properties) are based in the space–time interactions of processes with different “memory” (latency). We draw inspiration from the corresponding underlying biological mechanisms to propose two general schemes for directionality computation in nets, compatible with other space–time filtering properties. First, a connectivistic mechanism based on bipolar–amacrine–ganglion cell interaction is proposed, by formalizing the classical proposals of early vision neurophysiologists. Second, inspired initially in the more recently described intrinsic directionality of amacrines, novel schemes are proposed where directionality appear as the computing consequence of adding memory to spatial filtering structures. The mathematical formulations are achieved by means of Newton Filters and Hermite Functionals.

New biomimetic neural structures for artificial neural nets

The general aim is to formalize known properties of real neurons, formulating them into appropriate mathematical models. These will converge into, hopefully, more powerful neurophysiological founded distributed computation units of artificial neural nets. Redundancy and distributed computation are key factors to be embodied in the corresponding biomimetic structures. We focus in two neurophysiological processes: first, the dendro-dendritic or afferent non linear interactions, prior to the synapses with the cell body. Computational redundancy (and reliability as a consequence) is to be expected. Second, distributed computation, also provoked by a dendritic-like computational structure to generate arbitrary receptive fields weights or profiles, where also, a kind of reliability is expected, result of the distributed nature of the computation.

Local Space-Time Systems Simulation of Linear and Non-linear Retinal Processes

The realization of local space-time models of retinal processes can be achieved by means of available typical dynamical systems simulation tools (like Simulink) because only a very small number of parallel channels is needed. In short, the aim is to simulate both the time and space dimensions as delay chains, where the travelling signals are available at different points of the delay chain to interact among them. These models provide an interesting and fruitful insight into the neurophysiological processes.

McCulloch’s Relation to Connectionism and Artificial Intelligence

It is normally accepted that the beginnings of modern computing connectionism can be traced to McCulloch and Pitts’ paper of 1943 [1]. The important points of their historical contributions are however mislead by the drift that developments on theoretical computer architectures took after the 50’s. The so called Artificial Neural Nets and subsequent connectionist philosophy were actually fixed by Rosenblatt’s Perceptrons and his detractors, plus the more recent addenda of multi-layer perceptrons and back propagation adjusting techniques. They clearly used the basic idea of threshold logic and computation but evolved away from McCulloch-Pitts proposals, towards and in a computer tool of many times questionable power, just as parametric classifiers.

Eulerian Numbers Weigths in Distributed Computing Nets

We explore the possibilities of Eulerian numbers to define weights in layered networks and model distributed computation at the level of neurons receptive fields. These networks are then compared to those defined by binomial coefficients (Newton filters). Their potential as structures for signals convergence, divergence and overlapping is also established.

A Class of 3-D Distributed Modular Computing Nets

The class of flat triangular and layered nets of simple computing units, which gives rise to Newton and Hermite filters in two dimensions are extended to 3-D by means of two natural discrete ways. First, by means of the so called Pascal Pyramids; and second, by introducing a rectangular grilled “retina”, which leads to a kind of Newton quadrangular pyramids and to 3-D Newton Filters and Nets. Both cases can be extended to the continuous in the form of 2-D Hermite functions and filters of different orders (degree of derivatives). Preliminary results and examples are presented.

Control and Command Systems Concepts from Early Work on a Mars Rover

We recover and develop some robotic systems concepts (on the light of present systems tools) that were originated for an intended Mars Rover in the sixties of the last century at the Instrumentation Laboratory of MIT, where one of the authors was involved. The basic concepts came from the specifications for a type of generalized robot inspired in the structure of the vertebrate nervous systems, where the decision system was based in the structure and function of the Reticular Formation (RF). The vertebrate RF is supposed to commit the whole organism to one among various modes of behavior, so taking the decisions about the present overall task. That is, it is a kind of control and command system. In this concepts updating, the basic idea is that the RF comprises a set of computing units such that each computing module receives information only from a reduced part of the overall, little processed sensory inputs. Each computing unit is capable of both general diagnostics about overall input situations and of specialized diagnostics according to the values of a concrete subset of the input lines. Slave systems to this command and control computer, there are the sensors, the representations of external environment, structures for modeling and planning and finally, the effectors acting in the external world.

Remarks on neurocybernetics and its links to computing science. To the memory of Prof. Luigi M. Ricciardi.

This paper explores the origins and content of neurocybernetics and its links to artificial intelligence, computer science and knowledge engineering. Starting with three remarkable pieces of work, we center attention on a number of events that initiated and developed basic topics that are still nowadays a matter of research and inquire, from goal directed activity theories to circular causality and to reverberations and learning. Within this context, we pay tribute to the memory of Prof. Ricciardi documenting the importance of his contributions in the mathematics of brain, neural nets and neurophysiological models, computational simulations and techniques.

Analytical Models for Transient Responses in the Retina

We propose analytical models for slow potentials transients as they are recorded at different layers of the retina, but mostly as they are integrated to produce ganglion cells outputs. First, two possible pathways from photoreceptors to bipolars and to ganglia are formulated, their formal equivalence being shown for quasi-linear behaviour. Secondly, a linear oscillator analytical model is introduced which is shown also to be equivalent to the first under certain circumstances. Finally, local instantaneous nonlinearities are introduced in the models. Tunning their parameters, the models are very versatile in describing the different neurophysiological situations and responses.