arXiv subjects

Endel Poder

Publications and source records attributed to Endel Poder.

4 recordsLinked to original sources

arXiv2021

CNN-based search model underestimates attention guidance by simple visual features

Recently, Zhang et al. (2018) proposed an interesting model of attention guidance that uses visual features learnt by convolutional neural networks for object recognition. I adapted this model for search experiments with accuracy as the measure of performance. Simulation of our previously published feature and conjunction search experiments revealed that CNN-based search model considerably underestimates human attention guidance by simple visual features. A simple explanation is that the model has no bottom-up guidance of attention. Another view might be that standard CNNs do not learn features required for human-like attention guidance.

cs.CV↗

arXiv2020

On the rules of science game

Credit allocation in the mainstream bibliometrics is fundamentally flawed and the popular indicators have been misleading science for decades. Originally a simple technical mistake has become an integral part of our culture and is very difficult to correct. Although the problem has been raised in scientific articles, it seems mostly unknown to wider audience.

cs.DL↗

arXiv2017

Capacity limitations of visual search in deep convolutional neural networks

Deep convolutional neural networks follow roughly the architecture of biological visual systems and have shown a performance comparable to human observers in object recognition tasks. In this study, I tested three pretrained deep neural networks in visual search for simple visual features, and for feature configurations. The results reveal a qualitative difference from human performance. It appears that there is no clear difference between searches for simple features that pop out in experiments with humans, and for feature configurations that exhibit strict capacity limitations in human vision. Both types of stimuli reveal comparable capacity limitations in the neural networks tested here.

cs.CV↗

arXiv2015

A framework for the measurement and prediction of an individual scientist's performance

Quantitative bibliometric indicators are widely used to evaluate the performance of scientists. However, traditional indicators do not much rely on the analysis of the processes intended to measure and the practical goals of the measurement. In this study, I propose a simple framework to measure and predict an individual researcher's scientific performance that takes into account the main regularities of publication and citation processes and the requirements of practical tasks. Statistical properties of the new indicator - a scientist's personal impact rate - are illustrated by its application to a sample of Estonian researchers.

cs.DL↗