Welcome, Guest
Username: Password: Remember me
  • Page:
  • 1
  • 2

TOPIC: Papers aprobados

Papers aprobados 3 months 1 week ago #7696

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1580
  • Karma: 41
  • Thank you received: 144
Jose Arnoldo Segura

An Agent System for the analysis of Routine based Consumer Behavior


Marketing is a complex area that tries to understand the human mind throughout statistics and psychology. Their goal is to determined different ways to get to the consumer and make them try what they are selling. This goal proposes challenges as to what product a consumer might like to use, based on (not in a specific order) their beliefs, environment, culture, interest, availability, age, behavior, etc. In marketing, the understanding of consumer behavior is key as it is the starting point to determine by which way a product shall be launched? What is a good time to present a product? How much attention the campaign will get from the consumer? And what perception the consumer might get of the product. As of right now, marketing uses collectible data and statistical patterns in order to determine all those the answer to all previous questions, but in the field artificial intelligence, consumer behavior can be a problem that can be tackled, because patterns can be determined about the behavior of a particular type of consumer and multiagent systems can use those patterns in order to reproduce a similar behavior of the particular type of consumer that is being studied. Artificial Intelligence may need the statistics marketing is already using but it can add the missing component that the real consumer behavior is hiding: Rationing. Presented in 2002, Lamjed B., Drogoul A. and Bouron T. propose CUBES as a multiagent system based on observed behavior of individuals, socio-economic profiles and interaction with rules; and through simulation processes generate a community of consumers to see how they were influenced by the marketing campaigns. However, I propose an agent system that is based on the routines of the target market, consumers move themselves thru their routines and focus their attention and efforts to complete the task at hand in their routine. A well-defined target population share not only age, environment, and possibly culture, most certainly they share a routine that can be analyze as a pattern and through an agent system it can reproduced the target behavior and therefore their interest. Based on these two last items a product campaign can receive valuable information about the time (schedule ranges), place (social media, general media) in which the campaign will receive the maximum exposure alongside obtaining the maximum interest of the target audience.


Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,

Papers aprobados 3 months 1 week ago #7698

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1580
  • Karma: 41
  • Thank you received: 144
José Marín

A Reinforcement Learning approach to two-player stochastic games

Non-determinism and information incompleteness are, naturally, a serious concern in any decision taking process. An stochastic two-player game is a dynamic game with probabilistic transitions played by exactly two players. The players receive a payoff based on the actions they take. Each turn, the game moves to a new random state whose distribution depends on the current state and the chosen actions. This games define non-stationary environments in which both of the players are constantly learning and adapting. The traditional approach, in game theory, is strictly related to the Nash equilibrium. In other terms, the base policy assumes each of the players is playing optimally with respect to the other. Reinforcement learning techniques have already addressed similar problems in single agent stationary environments using Markov Decision Processes (MDPs). The goal is to test various Reinforcement Learning extensions of MDPs for multi-agent situations against real players of a two-player stochastic game. This extensions include value/policy iteration and Q-Learning variations. All of these following the underlying concept of Nash equilibrium. The experiments will be based on HearthStone, an online collectible card video game with high levels of probability, in terms of decision making and resource management, involved during gameplay.


Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,

Papers aprobados 3 months 1 week ago #7699

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1580
  • Karma: 41
  • Thank you received: 144
Jose Somarribas.

Alternative hardware optimizations for Virtual Reality applications ideal performance


Currently, Virtual Reality is becoming one of the top trending topics when talking about technology, from applications like Entertainment (games, shopping, tourism, etc.) and Medical to Military related ones, VR is now changing our lives and its use is growing at an exponential rate. As a consequence to its relatively recent exposure to the word, we are facing an era where VR is still not mature enough and where its public acceptance and acquisition has been negatively influenced by factors like high costs, possibility of motion sickness and lack of content among others. There is common factor between the three mentioned issues: Poor optimization. With an optimized VR technology costs will go down(less power consumption/less computation force needed), probability of motion sickness will decrease (higher frame rates and lower latency) and high quality content will also increase (more developers will start working on VR due to its cost effectiveness). VR optimization until now has been mainly driven by software developers trying to optimize its VR software in order to better take advantage of the existing hardware, but, is it possible to optimize the hardware(CPU, GPU, etc.) and not only the software in order to have a better VR experience and find its optimal performance? With the help of different software/hardware profiling tools, my proposal is to understand what currently the hardware VR bottlenecks are and come up with a quantifiable optimization in terms of hardware architecture, all this in order to reach nearly ideal Virtual Reality hardware/software performance.

Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,

Papers aprobados 3 months 1 week ago #7700

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1580
  • Karma: 41
  • Thank you received: 144
Michael Sánchez

Defining a specific sing language notation through machine learning

In Costa Rica, according to the INEC 2011 Census, there are 70709 people that have a hearing disability (with different scales of deafness) and they own the right to use the Costa Rican sing language (“LESCO” by its meaning in Spanish: lengua de señas costarricense) as their native language by law. Since this law was emitted in the 2012, the use of the LESCO has been increasing over the years and this is recognized in the deaf community as a positive accomplishment, but still, there are many things to improve and solve. This problem it is worth to aboard it because it includes the human factor in the computer science field, which it should be its main objective; besides, there could be a very large quantity of people that do not know LESCO, so it could produce a great negative impact on the Costa Rican society. In this matter, Serrato-Romero and Chacón-Rivas (2016) have been working a “Traductor LESCO: un esfuerzo puntual en el apoyo al proceso de aprendizaje de estudiantes con discapacidad auditiva”, this investigation aims to support the communication process in the channel hearing-deaf people; they use a web editor showed through a digital avatar to communicate information in Costa Rican sign language to deaf people in web sites and other environments. Nevertheless, there is the issue when the hearing people do not understand LESCO so it creates a communication barrier, that is why Quesada, Marín-Raventós and Guerrero (2016) are developing a “Sing Language Recognition Model Combining Non-Manual Markers and Handshapes” in order to positively assist this situation, so not only the hearing-deaf, but the deaf-hearing channel of communication could be helped. In these projects, still there is not a defined way to process the sign language grammar and all its morphological structures. The main goal of this paper is to investigate and find a way to integrate the sign language editor and the sign language recognition model to specified a notation for the sign language through machine learning; this way, it could be improved the transferring data speed of the LESCO computational analysis and provide a previous stage to parsing the LESCO in a computational environment.


Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,
  • Page:
  • 1
  • 2
  • Not Allowed: to create new topic.
  • Not Allowed: to reply.
  • Not Allowed: to add attachements.
  • Not Allowed: to edit your message.
Time to create page: 0.243 seconds
Powered by Kunena Forum