Machines that learn
https://www2.informatik.uni-hamburg.de/wtm/SocialRobotsWorkshop2016/images/display/assistive_2.jpg

Machines that learn

[en] Can a machine learn? Of course they can. There is a difference between programming and training a system. When you program, you tell the machine what to do in each situation. Faced with an unknown situation, the machine would not know what to do.

This methodology allows controlling the behavior of the system (maybe a robot, an Artificial Intelligence application or another). But it gets complicated when cases are many or when they are difficult to describe.

For example, a robot could find different situations in a restaurant, where chairs and dishes often change their places. A self-driving car could faces different drivers every time.

In such cases, Machine Learning techniques can be use. You don't program, rather you trains the system. What is the inputs? Just some "examples".

Examples of cases. Images of faces to learn who belongs to that company. Examples of positive texts for learning to recognize a good opinion from a negative one. Do you need to use a lot of examples? No, it depends on the case and the technique that you are applying.

There are several mathematical algorithms for Machine Learning, such as Neural Networks and Support Vector Machines. Among the Neural Networks, currently, it is growing the use of Deep Learning ones. Each one has its own characteristics and is used depending on the topic that you want to teach to the machine.


More info: What is that thing called "Deep Learning"? - https://www.dhirubhai.net/pulse/what-thing-called-deep-learning-marcela-riccillo

----------------------------------------------------------

[es] ?Puede una máquina aprender? Claro que puede. Hay una diferencia entre programar y entrenar un sistema. Cuando uno programa, le indica a la máquina qué hacer en cada situación. Ante una situación desconocida, la máquina no sabría qué hacer.

Esta metodología permite un control del comportamiento del sistema (sea un robot, una aplicación de Inteligencia Artificial u otro). Pero se complica cuando las situaciones pueden ser muchas o cuando son difíciles de describir.

Por ejemplo, las diferentes situaciones que puede encontrar un robot en un restaurant, donde las sillas y los platos cambian de lugar constantemente.O un auto robot que se enfrenta a diferentes conductores cada vez.

En esos casos pueden utilizarse técnicas de Machine Learning donde uno no programa, sino que entrena al sistema. ?Cuál es el input? Simplemente "ejemplos".

Ejemplos de casos. De imágenes de rostros para que aprenda cuáles personas pertenecen a una empresa. Ejemplos de textos positivos para que aprenda a distinguir una opinión buena de una negativa. ?Es necesario que sean muchos ejemplos? No, depende del caso y de la técnica.

Hay varios algoritmos matemáticos para Machine Learning, como Redes Neuronales y Máquinas de Soporte Vectorial. Entre las Redes neuronales, se destacan actualmente las de Deep Learning. Cada uno tiene sus características y se usan dependiendo el tema que se le ense?ará a la máquina.


Más información: ?Qué es esa cosa llamada "Deep Learning"? - https://www.dhirubhai.net/pulse/what-thing-called-deep-learning-marcela-riccillo

Picture source https://www2.informatik.uni-hamburg.de/wtm/SocialRobotsWorkshop2016/images/display/assistive_2.jpg

要查看或添加评论,请登录

Marcela Riccillo的更多文章

  • Trends 2017 - Robotics & Artificial Intelligence

    Trends 2017 - Robotics & Artificial Intelligence

    [en] A new year begins with renewed energy. And the new challenges begin.

    3 条评论
  • Humanoids in Hospitals

    Humanoids in Hospitals

    [en] There are a variety of applications of Robotics and Artificial Intelligence in Medicine. Starting with the Da…

  • What is in a robot hardware?

    What is in a robot hardware?

    [en] Currently, it is still not clear how to define a robot. It does not necessarily have to have humanoid form, it can…

  • Blues in Buenos Aires

    Blues in Buenos Aires

    [en] Walking the other day in Puerto Madero of Buenos Aires, it began to rain. And the image of the sky and the water…

  • NASA Robosimian exploded!!!

    NASA Robosimian exploded!!!

    [en] Robosimian robot was created by the NASA JPL Jet Propulsion Laboratory to participate in the DARPA Robotics…

  • I will be part of jury in a science TV Show!!!

    I will be part of jury in a science TV Show!!!

    [en] TV Channel Encuentro from Argentina will produce a science program called Eureka, and I will be part of the stable…

  • Toyota introduces a new little robot for travels - Kirobo Mini

    Toyota introduces a new little robot for travels - Kirobo Mini

    [en] The company Toyota is determined to achieve to make self-driving cars that could be robust cars and avoid…

    2 条评论
  • Google, IBM, Microsoft, Amazon & Facebook joined for AI

    Google, IBM, Microsoft, Amazon & Facebook joined for AI

    [en] As I have written in several posts, Artificial Intelligence is different from the Full Artificial Intelligence…

  • I Am A Visionary

    I Am A Visionary

    [en] In 2013 Google, with Andy Rubin being the Director of the Robotics Division, acquired several companies related…

    2 条评论
  • IBM helped to create the first cognitive movie trailer

    IBM helped to create the first cognitive movie trailer

    [en] IBM and 20th Century Fox teamed up to make the trailer of the thriller film called Morgan. The film is about a…

    1 条评论

社区洞察

其他会员也浏览了