Fundamentals of ML and AI
Machine Learning Fundamentals
Building an ML model means collecting and preparing data, choosing an algorithm, training, and then evaluating and iterating. A model is only as good as its data: garbage in, garbage out.
Training data
- Labeled data has a target value for each example, such as images tagged "cat" or "dog". Unlabeled data has only input features.
- Structured data is organized in rows and columns:
- Tabular: spreadsheets, databases, CSV files.
- Time series: values over time, such as stock prices, sensor readings, or weather.
- Unstructured data has no predefined format: text (documents, posts) and images (photos, video frames). It needs more advanced techniques.
Learning types
| Type | Data | Goal |
|---|---|---|
| Supervised | Labeled | Learn a mapping from inputs to known outputs, then predict outputs for new data |
| Unsupervised | Unlabeled | Discover hidden patterns, structure, or groupings |
| Reinforcement | Rewards and penalties from an environment | Learn actions that maximize reward over time through trial and error |
Semi-supervised learning trains on data where only a portion is labeled.
Inferencing
Inferencing is using a trained model to make predictions.
- Batch inferencing analyzes a large set of data at once. Use it when accuracy matters more than speed, as in data analysis.
- Real-time inferencing responds instantly to new data. Use it for chatbots and self-driving cars.