ECS170-L26

📗 -> 05/30/25: ECS170-L26

---

Machine Learning Slides

🎤 Vocab

❗ Unit and Larger Context

Small summary

✒️ -> Scratch Notes

Continuing from Unsupervised Learning

Unsupervised

Discover similar patterns, but no labels are available. However, there is a pattern in data
Clustering

• Images, Web results, …

Reinforcement

An agent (learner) interacts with an environment and watches the result of the interaction. Environment gives feedback via a positive or negative reward signal.

• No training data at the beginning
• Have to interact with the environment and learn from delayed feedback
  Playing chess…

Generalization

We care about the TEST error
Algorithms can can generalize
Similar to taking a midterm:

• Taking a practice midterm (training) is good, but we care about score on actual midterm (testing)
• Memorization vs. Learning

I.I.D Assumption

Independent and Identically Distributed (IID):

• All objects come from the same distribution (identically distributed)
• Objects are sampled independently (order doesn’t matter)
• We do NOT need to know the underlying distribution as long as the samples are sampled iid

In card terms:

• Pick a card, put it back in the deck, re-shuffle, repeat.
• Pick a card, put it back in the deck, repeat.
• Pick a card, don’t put it back, re-shuffle, repeat

Generating test data:

• Split data in two parts: (80/20 or 70/30):
  • First part train
  • Second part ONLY evaluates
  • Splitting should be random

Models

• Decision trees: A hierarchy of if-then-else rules (your classic “should you take an umbrella, or what decision should you take?)
• Linear models: . SVM for example
• Neural Networks

Complexity

The richness/complexity of a model increases with layers/depth, as the function space it can represent increases

• It can learn more complex decision rules
• Training error will reliably decrease. However, as you increase depth test error might increase with overfitting

🧪 -> Refresh the Info

Did you generally find the overall content understandable or compelling or relevant or not, and why, or which aspects of the reading were most novel or challenging for you and which aspects were most familiar or straightforward?)

Did a specific aspect of the reading raise questions for you or relate to other ideas and findings you’ve encountered, or are there other related issues you wish had been covered?)

🔗 -> Links

Resources

• Put useful links here

Connections

• Link all related words