Improving Deep Interactive Evolution with a Style-Based Generator for Artistic Expression and Creative Exploration.

Deep interactive evolution (DeepIE) combines the capacity of interactive evolutionary computation (IEC) to capture a user's preference with the domain-specific robustness of a trained generative adversarial network (GAN) generator, allowing the user to control the GAN output through evolutionary exploration of the latent space. However, the traditional GAN latent space presents feature entanglement, which limits the practicability of possible applications of DeepIE. In this paper, we implement DeepIE within a style-based generator from a StyleGAN model trained on the WikiArt dataset and propose StyleIE, a variation of DeepIE that takes advantage of the secondary disentangled latent space in the style-based generator. We performed two AB/BA crossover user tests that compared the performance of DeepIE against StyleIE for art generation. Self-rated evaluations of the performance were collected through a questionnaire. Findings from the tests suggest that StyleIE and DeepIE perform equally in tasks with open-ended goals with relaxed constraints, but StyleIE performs better in close-ended and more constrained tasks.

A Community Perspective on Resilience Analytics: A Visual Analysis of Community Mood.

Social networks are ubiquitous in everyday life. Although commonly analyzed from a perspective of individual interactions, social networks can provide insights about the collective behavior of a community. It has been shown that changes in the mood of social networks can be correlated to economic trends, public demonstrations, and political reactions, among others. In this work, we study community resilience in terms of the mood variations of the community. We have developed a method to characterize the mood steady-state of online social networks and to analyze how this steady-state is affected under certain perturbations or events that affect a community. We applied this method to study community behavior for three real social network situations, with promising results.

State and parameter estimation of a neural mass model from electrophysiological signals during the status epilepticus.

Status epilepticus is an emergency condition in patients with prolonged seizure or recurrent seizures without full recovery between them. The pathophysiological mechanisms of status epilepticus are not well established. With this argument, we use a computational modeling approach combined with in vivo electrophysiological data obtained from an experimental model of status epilepticus to infer about changes that may lead to a seizure. Special emphasis is done to analyze parameter changes during or after pilocarpine administration. A cubature Kalman filter is utilized to estimate parameters and states of the model in real time from the observed electrophysiological signals. It was observed that during basal activity (before pilocarpine administration) the parameters presented a standard deviation below 30% of the mean value, while during SE activity, the parameters presented variations larger than 200% of the mean value with respect to basal state. The ratio of excitation-inhibition, increased during SE activity by 80% with respect to the transition state, and reaches the lowest value during cessation. In addition, a progression between low and fast inhibitions before or during this condition was found. This method can be implemented in real time, which is particularly important for the design of stimulation devices that attempt to stop seizures. These changes in the parameters analyzed during seizure activity can lead to better understanding of the mechanisms of epilepsy and to improve its treatments.

An algorithm for on-line detection of high frequency oscillations related to epilepsy.

Recent studies suggest that the appearance of signals with high frequency oscillations components in specific regions of the brain is related to the incidence of epilepsy. These oscillations are in general small in amplitude and short in duration, making them difficult to identify. The analysis of these oscillations are particularly important in epilepsy and their study could lead to the development of better medical treatments. Therefore, the development of algorithms for detection of these high frequency oscillations is of great importance. In this work, a new algorithm for automatic detection of high frequency oscillations is presented. This algorithm uses approximate entropy and artificial neural networks to extract features in order to detect and classify high frequency components in electrophysiological signals. In contrast to the existing algorithms, the one proposed here is fast and accurate, and can be implemented on-line, thus reducing the time employed to analyze the experimental electrophysiological signals.