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

TOPIC: Papers aprobados

Papers aprobados 11 months 2 days ago #7690

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1810
  • Karma: 45
  • Thank you received: 170
En este thread aparecerán los papers aprobados. En otras palabras, mientras no sean mencionados acá no tienen tema aprobado.

-F.

Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,

Papers aprobados 11 months 2 days ago #7691

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1810
  • Karma: 45
  • Thank you received: 170
Iván Andrés Salazar Solano

Testing machine learning algorithms for self driving cars using video games as simulators

Nowadays we are living in a revolution where the self-driving cars are become main trend of technology automation. The advance in computer processing power and the research of artificial intelligence makes possible for a computer to simulate and react in matter of fraction of a second close to the thinking of a human. Also this increment in the processing capabilities have been used in the video game industry. No one can deny that the video games have become more realist each year. Each day, millions of people commute by car to their jobs and it becomes normal to want to automate a repetitive activity such as driving a car. Another objective with self driving cars is to prevent traffic accidents and traffic jams. Already car companies are implementing alerts that prevent the driver of getting outside the line or become closer to a object. But this alerts depends of the driver to react and make the decision. At this moment, the most important universities and companies around the world have projects and algorithms to make a car self driving. But in order to implement this algorithms in a car, the projects needs to build and test several scenarios whether in a simulation or in a controlled environment such as university campus or uninhabited town. This testing process becomes a problem because it results to be expensive in terms of time, human and economical resources to create a realistic simulator or adapt a car with sensors and instruments, hire actors for a real world test. For that reason, the idea is to use video games that simulates environments, like city or countryside with diversity of climates, so this algorithms can be safely tested and tuned. Also this idea enables to reproduce scenarios, that becomes difficult or risky to replicate in the real world, with the option to repeat them multiple times, adding random or programmed variables, without harming living beings and properties. This idea would allow to reduce the margin error and unexpected events as well.

Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,

Papers aprobados 11 months 2 days ago #7692

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1810
  • Karma: 45
  • Thank you received: 170
Rackel Elizondo

Rough Sets applied in the selection of attributes before a great dimensionality of the data


The decision to convict a suspect to prison is usually made deliberately and reliant on the judge handling the case. It is not uncommon that the decision is incorrect or fortunately, can be a correct decision. Rough sets is an excellent mathematical tool to obtain a safe and possible result based on a set of criteria and variables. When a person commits a misdemeanor, they often do not get convicted or punished for lack of budget to carry out their trial. Through the use of rough sets can be obtain a reliable result and determine if the suspect is guilty of their crime, it is also interesting because every time the variables and cases are recorded increase, more complex crimes could be evaluated, to in turn, the results are more neutral and as the dataset grows, more accurate results are obtained. In “Rough sets based on reducts of conditional attributes in medical classification of the diagnosis status” by Elisabeth Rakus-Andersson, rough sets are used for the evaluation of symptoms to obtain a medical diagnosis and also, in “A hybrid intelligent diagnosis approach for quick screening of Alzheimer's disease based on multiple neuropsychological rating scales” by Yin et al, rough sets have been used to evaluate diseases such as Alzheimer's.
The use of rough sets seeks to solve the problem of obtaining a fair, cheap and quick verdict on minor offenses. The social, psychological, and testimony of the suspects and their witnesses will help to obtain more readable and accurate results to make a decision of condemnation.

Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,

Papers aprobados 11 months 2 days ago #7693

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1810
  • Karma: 45
  • Thank you received: 170
Jennifer Caballero

Multiplicative LTSM with fast weights


For humans, visual recognition tasks seem easy, we effortlessly detect and classify objects from among tens of thousands of possibilities and we do so within a fraction of a second (DiCarlo et al., 2012). However, for computers, the object recognition problem is hard to solve. One machine learning method, deep learning is making major advances in solving problems that have resisted the best attempts of the artificial intelligence community for many years (LeCun et al., 2015), it has beaten records in image recognition tasks. Many technological advances depend on how well computers understand images, for instance, autonomous cars, UAV navigation, robotics and surveillance use camera sensors as a primary source of data from the world, improvements in the understanding of visual inputs will result in more reliable operation of these technologies. Typical recurrent neural networks with long short-term memory architecture (LTSM) use two types of memory, long-term that changes slow and short-term that changes fast. Recently, machine learning researchers have found that by using a third kind of memory implemented with fast weights, the performance of recurrent neural networks on different tasks can be improved.[1] Also researchers have combined LTSM with multiplicative recurrent neural networks architectures, outperforming typical LTSM and showed that this improvement increases as the complexity increases. I propose to combine the fast-weight third memory architecture with multiplicative recurrent neural network architecture in order to increase the performance of the typical LTSM and the LTSM with multiplicative architecture. Hopefully the results will show a better performance, while being able to keep that improvement as the complexity of the task increases.

Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,

Papers aprobados 11 months 2 days ago #7694

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1810
  • Karma: 45
  • Thank you received: 170
Ernesto Rivera

Viability of real time ray tracing in commodity hardware

The level of realism that can be achieved using the ray tracing technique compared to others is not comparable. Nonetheless, this higher level of realism comes with a price of performance cost, making it unsuitable for applications that require real time rendering. A performance efficient ray tracer that could run in real time in commodity hardware will open the opportunity to step up the detail and realism that is being used in current computer graphics applications. Several real time ray tracing engines have been developed with different degrees of success, but a true solution that is able to comply with the performance requirements is yet to appear. Currently, there is a new API for hardware accelerated rendering named Vulkan that has shown several performance improvements against others API's used in the current ray tracing solutions. I propose the development of a real time ray tracer using the Vulkan API and measure its performance against other initiatives to check the viability of developing an efficient real time ray tracer for commodity hardware.

Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,

Papers aprobados 11 months 2 days ago #7695

  • torres
  • torres's Avatar Topic Author
  • Offline
  • Profesor
  • Profesor
  • Posts: 1810
  • Karma: 45
  • Thank you received: 170
Guillermo Cornejo

A Polinomial-time Algorithm to Detect Imposible Vector Clocks Sets

The problem of Impossible Vector Clocks sets can be stated as «Given a set of vector timestamps, decide whether or no there is a distributed history containing them». We present a polynomial-time algorithm to decide if such history exists, under the hypotesis that there is a finite number of rules to detect violations of the vector clock rules. Efficiency evaluation of the algorithm shows it is possible to develop a real-time implementation for monitoring distributed systems.

Please Identificarse to join the conversation.

Dr. Francisco J. Torres-Rojas,
Last Edit: by torres.
  • 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: 0.176 seconds
Powered by Kunena Forum