arXiv subjects

João Dias

Publications and source records attributed to João Dias.

23 records · Page 2Linked to original sources

arXiv2020

The Influence of Reward on the Social Valence of Interactions

Throughout the years, social norms have been promoted as an informal enforcement mechanism for achieving beneficial collective outcomes. Among the most used methods to foster interactions, framing the context of a situation or setting in-game rules have shown strong results as mediators on how an individual interacts with their peers. Nevertheless, we found that there is a lack of research regarding the use of incentives such as scores to promote social interactions differing in valence. Weighing how incentives influence in-game behavior, we propose the use of rewards to promote interactions varying in valence, i.e. positive or negative, in a two-player scenario. To do so, we defined social valence as a continuous scale with two poles represented by Complicate and Help. Then, we performed user tests where participants where asked to play a game with two reward-based systems to test on whether the scoring system influenced the social interaction valence. The results indicate that the developed reward-based systems were able to foster interactions diverging in social valence scores, providing insights on how factors such as incentives overlap individual's established social norms. These findings empower game developers and designers with a low-cost and effective policy tool that is able to promote in-game behavior changes.

cs.HC↗

arXiv2019

Smooth representations of unit groups of split basic algebras over non-Archimedean local fields

We consider smooth representations of the unit group $G = \mathcal{A}^{\times}$ of a finite-dimensional split basic algebra $\mathcal{A}$ over a non-Archimedean local field. In particular, we prove a version of Gutkin's conjecture, namely, we prove that every irreducible smooth representation of $G$ is compactly induced by a one-dimensional representation of the unit group of some subalgebra of $\mathcal{A}$. We also discuss admissibility and unitarisability of smooth representations of G.

math.RT↗

arXiv2019

Multi-task Learning and Catastrophic Forgetting in Continual Reinforcement Learning

In this paper we investigate two hypothesis regarding the use of deep reinforcement learning in multiple tasks. The first hypothesis is driven by the question of whether a deep reinforcement learning algorithm, trained on two similar tasks, is able to outperform two single-task, individually trained algorithms, by more efficiently learning a new, similar task, that none of the three algorithms has encountered before. The second hypothesis is driven by the question of whether the same multi-task deep RL algorithm, trained on two similar tasks and augmented with elastic weight consolidation (EWC), is able to retain similar performance on the new task, as a similar algorithm without EWC, whilst being able to overcome catastrophic forgetting in the two previous tasks. We show that a multi-task Asynchronous Advantage Actor-Critic (GA3C) algorithm, trained on Space Invaders and Demon Attack, is in fact able to outperform two single-tasks GA3C versions, trained individually for each single-task, when evaluated on a new, third task, namely, Phoenix. We also show that, when training two trained multi-task GA3C algorithms on the third task, if one is augmented with EWC, it is not only able to achieve similar performance on the new task, but also capable of overcoming a substantial amount of catastrophic forgetting on the two previous tasks.

cs.LG↗

arXiv2019

Dynamic Social Interaction Mechanics CrossAnt

Nowadays, big effort is being put to study gamification and how game elements can be used to engage players. In this scope, we believe there is a growing need to explore the impact game mechanics have on the players' interactions and perception. This work focuses on the application of game mechanics to lead players to achieve certain types of social interaction (we named this type of mechanics social interaction mechanics). A word matching game called CrossAnt was modified so that it could dynamically generate different social interaction mechanics. These mechanics consisted in different key combinations needed to play the game and were aimed to promote what we think are three important types of social interactions: cooperation, competition and individual exploration. Our evaluation consisted on the execution of several sessions where two players interacted with the game for several levels and had to find for themselves how to perform the actions needed to succeed. While some of the levels required the input from both players in order to be completed, others could be completed by each player independently. Our results show that cooperation was perceived when both players had to intervene to perform the game actions. However, longer interactions may still be needed so that the other types of interactions are promoted.

cs.HC↗

arXiv2012

How Fama Went Wrong: Measures of Multivariate Kurtosis for the Identification of the Dynamics of a N-Dimensional Market

This paper investigates the common intuition suggesting that during crises the shape of the financial market clearly differentiates from that of random walk processes. In this sense, it challenges the analysis of the nature of financial markets proposed by Fama and his associates. For this, a geometric approach is proposed in order to define the patterns of change of the market and a measure of multivariate kurtosis is used in order to test deviations from multinormality. The emergence of crises can be measured in this framework, using all the available information about the returns of the stocks under consideration and not only the index representing the market.

q-fin.ST↗