Introduction à l'IA et la représentation des connaissances

-

Pourquoi optons-nous à  l'IA ?

  • Les connaissances d'un expert sont subjectives et difficiles à formaliser.
  • Les formalismes utilisés pour la représentation des connaissances ne permettent pas un bon niveau d'abstraction et restent liés à l'implantation.
  • Résoudre un problème recrurent
  • Un problème qui ne se résout pas par un algorithme, car un algorithme résout un problème qu'on connais déjà sa solutution
Comment juge-ton un systéme IA ?
La qualité d’un système intelligent est celle de sa base de connaissance
De quoi il est constitué un système IA?
Dans un système intelligent on a 3 composants

  1. Base de connaissance 
  2. Une partie pour faire les inférences (raisonnement) 
  3. Une structure de contrôle pour orienter le raisonnement
Plateforme IA Microsoft :
Principalement il existe 4 plateforme Microsoft pour résoudre un problème en IA
  1. Cognitive Service :Permet de réagir intelligemment avec l'utilisateur, à travers un ensemble d'API pour détecter principalement le visage, le parole ou langage
  2. Bot service : Permet  votre application de communiqué directement avec votre utilisateur (Messanger , Slack , Cortana)
  3. Deep learning : c'est une plateforme d'apprentissage profonde, vous pouvez faire apprendre votre modèle de donnée afin d'améliorer la réponse du système   
  4. Analyse des données : une fois vous avez "Trainner" votre système il est temps de pouvoir analyser votre base de connaissance a fin de tirer des informations utile 

Commentaires

Enregistrer un commentaire

Posts les plus consultés de ce blog

Cross-Origin Resource Sharing and Azure Machine Learning web services

-
Image
Cross-Origin Resource Sharing (CORS) is a mechanism that allows web objects (e.g., JavaScript) from one domain to request objects (web service requests) from another domain. Cross-origin requests initiated from scripts have been subject to restrictions primarily for security reasons. CORS provides a way for web servers to support cross-site access controls, which enable secure cross-site data transfers. At the time of this writing, it should be noted that Azure Machine Learning web services do not currently support CORS requests. See the following link for more information: https://social.msdn.microsoft.com/Forums/en-US/b6ddeb77-30e1-45b2-b7c1-eb4492142c0a/azure-ml-published-web-services-cross-origin-requests?forum=MachineLearning. This CORS restriction really means that if you wish to fully exploit Azure Machine Learning web services for deployment, testing, and production for a wide variety of (web) clients, you will need to host your own server-side applications. You basically hav
Lire la suite

Azure Machine Learning Workflow

-
Image
The process of creating a Machine Learning on Azure  is composed of a many pattern of workflow steps. This workflow are designed to help users to create a new predictive alanytics in no limited time. The main steps in the process are summarized in this figure : Data :  Is your input, who will be acquired, compiled , analyzed, tested and trained Create the model :  There is a versious algorithms of machine learning. we should create a models who is capable to make predictions, based on interfaces about the datasets. Evaluate the model :  This step is so important in the process because we examine the accuracy of new predictive models based on ability to predict the correct outcome, when we know in advance the input and the output values. Accuracy is measured in terms of confidence factor approching the whole number one Refine and evaluate the model :  After evaluating the model we refine-it by comparing, contrasting and combining alternate predictive models to find th
Lire la suite

Naive Bayes Classification Algorithm

-
Image
You are a data scientist,  an Intelligence Artificielle engineer, or a Statisticien, this post can help you to resolve your problems You are working on a classification problem and you have generated your set of hypothesis, created features and discussed the importance of variables. Within an hour, stakeholders want to see the first cut of the model. What will you do? You have hunderds of thousands of data points and quite a few variables in your training data set. In such situation, if I were at your place, I would have used ‘Naive Bayes‘, which can be extremely fast relative to other classification algorithms. It works on Bayes theorem of probability to predict the class of unknown data set. You wold like to know haw it work ? Step 1: Convert the data set into a frequency table Step 2: Create Likelihood table by finding the probabilities like Overcast probability = 0.29 and probability of playing is 0.64. Step 3: Naive Bayesian equation to calculate the post
Lire la suite