Archives for 2005

Today I read a paper titled “Interactive visualization of higher dimensional data in a multiview environment”

The abstract is:
We develop multiple view visualization of higher dimensional data.

Our work was chiefly motivated by the need to extract insight from four dimensional Quantum Chromodynamic (QCD) data.

We develop visualization where multiple views, generally views of 3D projections or slices of a higher dimensional data, are tightly coupled not only by their specific order but also by a view synchronizing interaction style, and an internally defined interaction language.

The tight coupling of the different views allows a fast and well-coordinated exploration of the data.

In particular, the visualization allowed us to easily make consistency checks of the 4D QCD data and to infer the correctness of particle properties calculations.

The software developed was also successfully applied in material studies, in particular studies of meteorite properties.

Our implementation uses the VTK API.

To handle a large number of views (slices/projections) and to still maintain good resolution, we use IBM T221 display (3840 X 2400 pixels).

Today I read a paper titled “Neural network ensembles: Evaluation of aggregation algorithms”

The abstract is:
Ensembles of artificial neural networks show improved generalization capabilities that outperform those of single networks.

However, for aggregation to be effective, the individual networks must be as accurate and diverse as possible.

An important problem is, then, how to tune the aggregate members in order to have an optimal compromise between these two conflicting conditions.

We present here an extensive evaluation of several algorithms for ensemble construction, including new proposals and comparing them with standard methods in the literature.

We also discuss a potential problem with sequential aggregation algorithms: the non-frequent but damaging selection through their heuristics of particularly bad ensemble members.

We introduce modified algorithms that cope with this problem by allowing individual weighting of aggregate members.

Our algorithms and their weighted modifications are favorably tested against other methods in the literature, producing a sensible improvement in performance on most of the standard statistical databases used as benchmarks.

Today I read a paper titled “On the suitability of the 2 x 2 games for studying reciprocal cooperation and kin selection”

The abstract is:
The 2 x 2 games, in particular the Prisoner’s Dilemma, have been extensively used in studies into reciprocal cooperation and, to a lesser extent, kin selection.

This paper examines the suitability of the 2 x 2 games for modelling the evolution of cooperation through reciprocation and kin selection.

This examination is not restricted to the Prisoner’s Dilemma, but includes the other non-trivial symmetric 2 x 2 games.

We show that the popularity of the Prisoner’s Dilemma for modelling social and biotic interaction is justified by its superiority according to these criteria.

Indeed, the Prisoner’s Dilemma is unique in providing the simplest support for reciprocal cooperation, and additive kin-selected altruism.

However, care is still required in choosing the particular Prisoner’s Dilemma payoff matrix to use.

This paper reviews the impact of non-linear payoffs for the application of Hamilton’s rule to typical altruistic interactions, and derives new results for cases in which the roles of potential altruist and beneficiary are separated.

In doing so we find the same equilibrium condition holds in continuous games between relatives, and in discrete games with roles.

Today I read a paper titled “Finding Approximate Palindromes in Strings Quickly and Simply”

The abstract is:
Described are two algorithms to find long approximate palindromes in a string, for example a DNA sequence.

A simple algorithm requires O(n)-space and almost always runs in $O(k.n)$-time where n is the length of the string and k is the number of “errors” allowed in the palindrome.

Its worst-case time-complexity is $O(n^2)$ but this does not occur with real biological sequences.

A more complex algorithm guarantees $O(k.n)$ worst-case time complexity.

Today I read a paper titled “Intrusion Detection Systems Using Adaptive Regression Splines”

The abstract is:
Past few years have witnessed a growing recognition of intelligent techniques for the construction of efficient and reliable intrusion detection systems.

Due to increasing incidents of cyber attacks, building effective intrusion detection systems (IDS) are essential for protecting information systems security, and yet it remains an elusive goal and a great challenge.

In this paper, we report a performance analysis between Multivariate Adaptive Regression Splines (MARS), neural networks and support vector machines.

The MARS procedure builds flexible regression models by fitting separate splines to distinct intervals of the predictor variables.

A brief comparison of different neural network learning algorithms is also given.

Today I read a paper titled “Finding Traitors in Secure Networks Using Byzantine Agreements”

The abstract is:
Secure networks rely upon players to maintain security and reliability.

However not every player can be assumed to have total loyalty and one must use methods to uncover traitors in such networks.

We use the original concept of the Byzantine Generals Problem by Lamport, and the more formal Byzantine Agreement describe by Linial, to nd traitors in secure networks.

By applying general fault-tolerance methods to develop a more formal design of secure networks we are able to uncover traitors amongst a group of players.

We also propose methods to integrate this system with insecure channels.

This new resiliency can be applied to broadcast and peer-to-peer secure communication systems where agents may be traitors or become unreliable due to faults.