Genetic Algorithms In Complex Problem Space:
This paper describes a learning scenario, where a creature with tree legs learns to walk.
The aim of this project is to see the genetic algorithm converge in a problem domain of high complexity.

To produce a learning scenario of suitable complexity, we use a physics engine to simulate a 3d world, in which we place a learning agent with 3 legs with 2 joints each. Movement becomes a functional of the functions of each limb, with respect to speed, gravity, weight, position, and rotation.

The Tripod generates its actions from definitions expressed in its own gene structure.

Dowload Paper

The Sigma Pi Spiking Neuron:
This paper presents the “Sigma Pi Spiking Neuron”,
a novel biologically plausible computational model of a spiking neuron with
multiplicative response and multiplexed spike encoding.

The paper describes major computational dynamics at play in spiking neurons,
reproduces bio-plausible spike trains from simulated retinal ganglion neurons,
and show how to combine leaky-integrate-and-fire neurons with active dendrites,
to model bio-plausible contrast sensitivity.

Dowload Paper

Diversity of Evolutionary Algorithms in Data mining:
Evolutionary algorithms are well known for their brute problem solving force in vast and complex problem domains, and their capability of working without supervision.
Evolutionary algorithms produce good results, but often of a
undecipherable “black box” type. Due to intricate problem-domain dependent challenges,
and indecipherability issues, appliance to data mining is not trivial.
This paper reviews the work of pioneers, who have identified
areas in data mining where evolutionary algorithms are applicable.
Dowload Paper

Intricacies of the visual cortex:
This paper describes the neural network system paradigms of the visual cortex,
explain the particulars of the methodology, and show how:
Hetero associative networks,Auto associative attractor networks, Competitive networks,
Composite networks,Continuous attractor networks, and Strange attractor networks;
Constitutes the foundations of: Clustering, Generalization, Transference,
Inductive reasoning, Learning, Memory, and Modelability.
The paper further investigates the dynamics of
the visual cortex by modeling it as an “excitable media”.
The simulation visualizes the dynamics of 100 000 interacting neurons
by an OpenGl simulation written in C++.
Dowload Paper

Object recognition with neural networks:
This paper describes an implementation of object recognition with neural networks.
It outlines a process that achieves 100% object recognition in the task of
classifying 223 different pictures of fruit.

The recognition is achieved without use of any colour information except that of total saturation levels.
The pre-processing uses RGB level thresholding to rid the pictures of the background
and the reoccurrence of an occluding hand.

The approach deals with pictures of varying histogram density distributions,
and objects of varying scale, translation, and rotation.
Dowload Paper

Neural Network Time Series Prediction:
This paper describes neural network time series prediction project,
applied to forecasting the American S&P 500 stock index.
679 weeks of raw data is preprocessed and used to train a neural network.

The project is built with Matlab (Mathworks inc.)
Matlab is used for processing and preprocessing the data.
A prediction error of 0.00446 (mean squared error) is achieved.

On of the major goals of the project is to visualize how the network
adapts to the real index course by approximation
Dowload Paper

Visual Data Exploration Using Self Organization:
Clustering of multivariate data divides the data into groups, often previously unknown to the researcher, revealing important and potentially novel information on the structure in the data.

Clustering is especially meaningful for vast datasets with many variables,
where the complexity alone presents a comprehension barrier.
The SOM algorithm reduce the dimensionality of the data by mapping multivariate vectors to clusters on a 2 dimensional grid, not loosing information but defocusing on non-vital stochasticity , focusing on the higher-level patterns of vector distribution.

Dowload Paper