Short answer
Machine learning trains models on data so they improve at a task without being fully hardcoded—powering recommendations, fraud detection, speech systems, and many modern AI products.
Why it matters
If you understand ML basics—data, loss, generalization—you can ask better questions about accuracy, bias, and whether a vendor demo will survive contact with your real inputs.
Article
Study
Work through train/val/test splits, bias-variance intuition, and baseline models before deep nets.
Examples
Credit risk scoring with tabular models
Lenders often train gradient-boosted trees on historical application and repayment data to estimate default risk. Features may include income ratios, credit history summaries, and application metadata. Fairness audits, explainability reports, and human underwriting policies are typically required because errors have significant consequences.
Demand forecasting
Retailers forecast product demand using time-aware ML models that combine past sales, promotions, seasonality, and external signals. The forecast then drives inventory decisions. Evaluation focuses not only on average error but on costly misses during peak periods.
FAQ
What is the difference between training and inference?
Training is the process of fitting model parameters using data and an optimization procedure. Inference is using a trained model to produce predictions or generations on new inputs, usually without updating those parameters.
Supervised vs unsupervised learning — which should I use?
Use supervised learning when you have reliable labels for the outcomes you care about (spam/not spam, disease present/absent). Use unsupervised methods to explore structure, cluster entities, or compress data when labels are scarce. Many modern systems combine both, especially via self-supervised pretraining followed by supervised fine-tuning.
Related glossary terms
- Artificial intelligence — Computer systems designed to perform tasks that typically require human cognitive abilities, such as perception…
- Deep learning — Machine learning with multi-layer neural networks that learn hierarchical representations from large datasets.
- Embedding — A numeric vector representation of an item—such as a token, sentence, image, or document—used to capture similarity and…
- Fine-tuning — Additional training that adapts a pretrained model to a narrower task, domain, or preferred behavior using specialized…
- Inference — The process of running a trained model on new inputs to produce predictions or generations, usually without updating…
- Machine learning — A family of methods that improve model performance by learning patterns from data instead of relying only on hand-coded…
- Reinforcement learning — Learning to choose actions in an environment to maximize cumulative reward through interaction rather than fixed labels.
- Training data — The examples used to fit a model’s parameters, including inputs and any labels or supervisory signals required by the…
Sources
-
CS229: Machine Learning course materials
Stanford UniversityCanonical university course notes covering supervised learning, learning theory foundations, and practical ML methods.
-
Machine Learning Crash Course
GoogleFree practical curriculum on core ML concepts, generalization, and workflows.
-
DeepLearning.AI blog
DeepLearning.AIEducational articles connecting classical ML ideas to modern deep learning practice.