Artificial Intelligence and Machine Learning on Azure

Discover Azure’s main AI and ML services with practical examples.

Azure Machine Learning

Azure Machine Learning is a comprehensive platform for developing and deploying ML models.

Environment Setup

from azureml.core import Workspace, Experiment, Environment

# Create a workspace
ws = Workspace.create(
    name='mon-workspace',
    subscription_id='<subscription_id>',
    resource_group='mon-groupe'
)

# Create an environment
env = Environment.from_conda_specification(
    name='mon-env',
    file_path='environment.yml'
)

# Create an experiment
experiment = Experiment(workspace=ws, name='mon-experiment')

Training a Model

from azureml.core import ScriptRunConfig

# Training configuration
script_config = ScriptRunConfig(
    source_directory='./scripts',
    script='train.py',
    environment=env,
    compute_target='mon-cluster'
)

# Start training
run = experiment.submit(script_config)

Azure Computer Vision

Image analysis and computer vision service.

Image Analysis

from azure.cognitiveservices.vision.computervision import ComputerVisionClient
from msrest.authentication import CognitiveServicesCredentials

# Create the client
client = ComputerVisionClient(
    endpoint="<endpoint>",
    credentials=CognitiveServicesCredentials("<key>")
)

# Analyze an image
with open("image.jpg", "rb") as image:
    result = client.analyze_image_in_stream(
        image,
        visual_features=["Categories", "Description", "Tags"]
    )

# Display the results
print(f"Description: {result.description.captions[0].text}")
print("Tags:", ", ".join([tag.name for tag in result.tags]))

Azure Text Analytics

Text analysis and natural language processing service.

Sentiment Analysis

from azure.ai.textanalytics import TextAnalyticsClient
from azure.core.credentials import AzureKeyCredential

# Create the client
client = TextAnalyticsClient(
    endpoint="<endpoint>",
    credential=AzureKeyCredential("<key>")
)

# Analyze sentiment
documents = ["J'adore ce produit !", "Ce service est terrible."]
result = client.analyze_sentiment(documents)

# Display the results
for doc in result:
    print(f"Sentiment: {doc.sentiment}")
    print(f"Confiance: {doc.confidence_scores.positive}")

Azure Speech Services

Speech recognition and speech synthesis service.

Speech Synthesis

from azure.cognitiveservices.speech import SpeechConfig, SpeechSynthesizer

# Configuration
speech_config = SpeechConfig(
    subscription="<key>",
    region="<region>"
)
speech_config.speech_synthesis_voice_name = "fr-FR-HenriNeural"

# Create the synthesizer
synthesizer = SpeechSynthesizer(speech_config=speech_config)

# Synthesize speech
result = synthesizer.speak_text_async("Bonjour, je suis un assistant virtuel.").get()

# Check the result
if result.reason == ResultReason.SynthesizingAudioCompleted:
    print("Synthèse vocale réussie")

Azure Personalizer

Personalized recommendations service.

Configuring Recommendations

from azure.cognitiveservices.personalizer import PersonalizerClient
from azure.cognitiveservices.personalizer.models import RankRequest

# Create the client
client = PersonalizerClient(
    endpoint="<endpoint>",
    credential=AzureKeyCredential("<key>")
)

# Prepare the request
rank_request = RankRequest(
    actions=[
        {"id": "action1", "features": [{"category": "type", "value": "film"}]},
        {"id": "action2", "features": [{"category": "type", "value": "série"}]}
    ],
    context_features=[
        {"category": "user", "value": "jeune"},
        {"category": "heure", "value": "soir"}
    ]
)

# Get a recommendation
rank_response = client.rank(rank_request)
print(f"Action recommandée: {rank_response.reward_action_id}")

Best Practices

1. Cost Management

   • Use request quotas
   • Monitor usage
   • Optimize models

2. Performance

   • Cache results
   • Use batch processing
   • Optimize requests

3. Security

   • Manage API keys
   • Implement authentication
   • Protect data

4. Monitoring

   • Track metrics
   • Configure alerts
   • Analyze logs

Conclusion

Key takeaways:

• Comprehensive ML services
• Easy-to-use APIs
• Good documentation
• Integration with Azure

Recommendations:

• Start with the basic services
• Test with simple data
• Optimize costs
• Follow best practices
• Train on Microsoft Learn