Cada vez hay más librerías, plataformas y servicios en la nube para crear aplicaciones de Inteligencia Artificial (AI). Pero utilizarlas directamente para programar tu componente inteligente te obliga a aprender el lenguaje específico de cada plataforma y te liga a ella en el futuro. No es una buena thought con lo rápido que evoluciona el ecosistema de la IA con nuevas, y mejores, plataformas apareciendo a diario.
¿Como puedes crear una app que integre algún algoritmo de IA (lo que también se llama sensible apps ) incluso una solución completa de ciencia de datos (data science) sin tener que aprender a programar y depender de una plataforma concreta? The solution is to model the AI application and then let the modeling platform generate and run the application using, for example, some of the existing deep learning frameworks.
You should already know that by raising the level of abstraction at which you define your software problem, you immediately gain a series of benefits, such as technological independence. This is also the trend followed in the world of artificial intelligence. There are more and more tools for modeling AI to make it easier for everyone, even if they are not programmers, to access this fascinating technology.
Vamos a intentar cubrir tanto las herramientas que permiten añadir componentes de inteligencia artificial predefinidos a una aplicación existente como herramientas específicas para preparar y entrenar modelos de aprendizaje automático (machine learning).
Herramientas de modelado visual de IA de la mano de las grandes tecnológicas
Las grandes del mundo tech” (Google, Amazon, Microsoft) han visto los entornos de modelado para IA como una manera de crecer su base de usuarios. Te venden la programación visible” como manera de crear tus propias aplicaciones de aprendizaje automático sin programar pero, evidently, luego esas aplicaciones sólo las puedes ejecutar en sus plataformas respectivas.
Azure Machine Learning Studio
The modeling environment for Machine Learning that I like the most is Microsoft's Azure Machine Learning Studio. As Microsoft says: it is an easy environment, runnable in your browser and offering a visual and drag-and-drop environment where there is no need to write any kind of code. You can easily define what your input data is, process it (if necessary), use it to train different types of machine learning models and finally evaluate the quality of the results.
Creating a model for classification with Azure ML Studio
IBM's SPSS Modeler
IBM's alternative to Azure ML Studio is SPSS Modeler, part of Watson Studio, like Microsoft's solution, allows you to define your knowledge pipeline, the model you want to generate (classifier, predictive,...) and evaluate and visualize the results. It comes with a very complete library of predefined algorithms and models so you don't have to start from scratch.
The closest thing that Amazon provides would be Amazon SageMaker, but I haven't been able to see that it includes any type of visual editor for AI.
Modeling for data science
In a knowledge science problem, data collection and its manipulation/preparation is as important as the machine learning we are going to try next. That is why visual data science environments come with a large number of components for 'data massage'.
Although some of these tools may come with their own execution engine, la mayoría se integran con frameworks de deep studying existentes como Keras Tensorflow
RapidMiner
RapidMiner incluye un herramienta de diseño visual de workflows para prototipar y validar modelos predictivos. Viene con un buen número de conexiones predefinidas con servicios artificial intelligence blog externos (muchas de ellas para la integración de datos, RapidMiner soporta más de 60 tipos de ficheros y formatos de datos, tanto para datos estructurados como no estructurados).
Orange
Orange es una herramienta de aprendizaje y visualización de modelos de ML. El análisis de los datos se realiza vía la conexión de widgets en un flujo de datos común. Cada widget se encarga de una más tareas de recuperación, preproceso, visualización evaluación de datos. Almost everything you can imagine already has its widget, but you can still create your own.
Visual programming of workflows for data analysis with Orange
Knime
My favorite option. Knime is a generic platform for data analysis that can be used for many different purposes. There are more than 2000 types of nodes you can use for this. The Knime extensions for knowledge scientists and Knime for deep learning are the ones that interest us most for this submission. For example, The deep learning extension allows users to perform all kinds of operations (from definition to execution) on neural networks. It can be complemented with the extension for knowledge scientists to greatly enrich the whole data collection and processing part. In addition, Knime is open source
Notice that apart from these specific tools for knowledge science, we will increasingly see more and more extensions that add a certain level of data analysis functionalities in more general environments. A great example would be Neuron , an information science extension for Visual Studio Code
Modeling neural networks
If your main goal is neural networks, DIANNE is a good option. In DIANNE, neural networks are defined as a directed graph that can be visually created with its online editor from predefined modules.
If what you want is mainly to learn how neural networks work, then this TensorFlow playground is the very best solution. It allows you to play with neural networks (añadiendo / quitando neuronas y modificando sus parámetros) y aprender así sus conceptos básicos.
Visualización de los modelos de aprendizaje
Hay herramientas especializadas en visualizar los resultados del aprendizaje automático. Su objetivo es ayudarte a entender como funciona y como de bien responde el modelo de ML generado.
Herramientas de generación de código para crear software program con comportamiento inteligente
Las herramientas low-code clásicas” se están dando cuenta que sus usuarios quieren poder añadir componentes inteligentes a sus modelos software, de la misma forma que pueden diseñar el modelo de datos de comportamiento. For example, they want to be able to easily add chatbots to their software, a facial recognition component, without having to learn to program with deep learning libraries.
As a response, these code generation tools are already adding new visual modules that allow representing and encapsulating these intelligent functionalities. For now, only the integration of already predefined components is offered (to recognize objects, sentiment analysis,...) and offered by external platforms, so the integration basically consists of linking your software program with that external service with few configuration possibilities. But in any case, it is a good start.
Genexus
Genexus is the one that offers the most and best options, how to summarize the table that I copy below for you (notice that for some of the modules you can also choose on which infrastructure to run them):
Google Cloud AI
✓
✓
Genexus has recently also released an extension to generate chatbots and in fact we are looking at how it could be integrated with Jarvis, our own framework for modeling and deploying cross-platform chatbots.
Mendix
Mendix is moving in a similar direction (e.g.. see how to create a chatbot with Mendix ) but for now the process seems to require quite a bit more work guide.
Lobe
Lobe (now acquired by Microsoft) es un nuevo competidor en la area de las herramientas ” low-code para AI ” Su objetivo es que te sea muy fácil entrenar modelos de deep studying que luego puedas integrar en una app. Todavía están en beta y con la compra por Microsoft es difícil saber como van a evolucionar, pero viendo los ejemplos visuales de deep studying, es fácil imagina un buen número de apps interesante para dispositivos móviles que se podrían crear con Lobe.
Reconocimiento de gestos con Lobe
Related Posts
January 24, 2017
COMMENT
Identify
Email
Web site
Anúnciate aquí
Con más de 800 visitantes al día, somos la net líder en Ingeniería de Software program de habla hispana. Contacta para anunciarte antes de que sea demasiado tarde
Crea tus propios chatbots con Xatkit
Únete a nuestro grupo de Ingenieros Software program
No te pierdas las últimas novedades sobre programación, design, AI, metodologías y mucho más
Apúntate
Pin It on Pinterest
Digital platform on the technological uniqueness of Roko's Basilisk, We promote the artificial intelligence of the future.