Azure Machine Learning Training Hub

This training hub covers Azure Machine Learning from fundamentals to production operations, with formulas, architecture, deployment guidance, and visual references.

Progression model:

• Beginner: understand AI/ML fundamentals and platform basics.
• Intermediate: build data, training, and evaluation workflows.
• Advanced: deploy, monitor, debug, and govern production ML systems.

Evolution of Machine Learning

Machine learning as a scientific field started in the 1950s, then moved through multiple eras based on compute power, data availability, and algorithmic advances.

Note - How to read this timeline: Each colored band is the dominant paradigm of an era, and the dots on the axis are the breakthroughs that triggered the next shift. Read it to see why modern Azure ML must support both classic ML (still best for tabular business data) and deep/foundation models (best for unstructured data) : the platform spans the whole history, not just the latest era.

Milestone eras

1. Foundations (1950s-1970s): Turing test ideas, perceptrons, nearest-neighbor methods.
2. Expert systems era (1980s): rule-based AI in enterprise workflows.
3. Statistical ML era (1990s-2000s): SVMs, random forests, probabilistic modeling.
4. Deep learning era (2010s): GPUs and large datasets enabled deep neural networks.
5. Foundation model era (2020s+): transformers, large language models, multimodal AI.

Why this matters for Azure ML learners

• It explains why modern MLOps includes both classic ML and deep learning workflows.
• It clarifies when simpler models can outperform larger neural models in tabular data.
• It frames current production needs: monitoring, governance, safety, and cost control.

Learning Path

Module 01

Math Prerequisites

Probability, linear algebra, calculus, and statistics foundations needed for all ML math.

Module 02

End-to-End ML Overview

A map of the whole ML workflow and how every stage connects, from problem framing to production monitoring.

Module 03

Introduction and Lifecycle

AI vs ML vs data science, AI categories, and end-to-end Azure ML lifecycle.

Module 04

ML Foundations

Full ML taxonomy: supervised, unsupervised, RL, semi/self-supervised, with formulas and selection guidance.

Module 05

Neural Networks and Deep Learning

Perceptron to Transformers: backpropagation, CNNs, RNNs, attention, and training at scale.

Module 06

Azure ML Environment

Workspace taxonomy, compute types, model registry, and endpoints.

Module 07

Environment Setup

Conda/pip setup, package validation, and runtime consistency.

Module 08

Data Preparation

Data collection, cleaning, schema handling, and split strategy.

Module 09

Model Types

Algorithm families with representative mathematical formulations.

Module 10

Training and AutoML

AutoML search flow, compute choices, and practical training pipeline.

Module 11

Performance Metrics

Classification and regression metrics, formulas, and interpretation.

Module 12

Results and Explainability

Result analysis, drift detection, and explainability methods.

Module 13

Deployment

Registration, scoring, endpoint deployment, and serving patterns.

Module 14

Deployment Debugging

Kubernetes-focused troubleshooting for production endpoint issues.

Reference

Ref 01

Reference Home

Supporting material for implementation and operations.

Ref 02

Mathematical Model Deep Dive

Formula-level explanations, objectives, assumptions, and trade-offs for core ML models.

Ref 03

CLI Commands

Command-line references for setup, run, and deployment tasks.

Ref 04

Glossary

Core Azure ML and MLOps terms used throughout this training.