Machine Learning
Machine learning studies systems whose behavior is fitted from data or interaction rather than fully specified by hand. The algorithm is only one part of the system: the question, data-generating process, objective, evaluation design, and deployment conditions determine whether the result is useful.
Problem Families
| Family | Learning signal | Typical question |
|---|---|---|
| Supervised learning | paired inputs and targets | predict an outcome for a new case |
| Unsupervised learning | observations without task labels | discover structure or a useful representation |
| Self-supervised learning | targets constructed from the data itself | learn reusable representations or conditional predictions |
| Reinforcement learning | actions, state transitions, and rewards | choose behavior that improves long-run return |
These families can overlap. A deep neural network describes a model class, not a separate source of supervision.
Working Sequence
- Define the decision or prediction target and an explicit baseline.
- Inspect collection, sampling, labels, missingness, and possible leakage.
- Choose a loss that represents the training objective and metrics that represent actual use.
- Fit only on training data; use validation data for choices and hold the test set for final evaluation.
- Compare errors across relevant groups, time periods, and operating conditions.
- Monitor distribution shift, feedback loops, privacy, and failure cost after deployment.
Site Path
- General Concepts introduces the shared vocabulary.
- Linear Models provides interpretable regression and classification baselines.
- Deep Learning develops representation learning and neural architectures.
- Data Science covers the wider evidence and reproducibility loop.
Berkeley CS 189 is the primary external course path. Product services, current model APIs, and benchmark snapshots belong elsewhere because they change faster than this foundation.