WorksheetsA Few Useful Things About Machine Learning
Total questions: 12
Worksheet time: 6mins
Which of these is an example of overfitting in real life?
Memorizing every question from last year’s exam
Practicing general problem-solving strategies
Learning to cook with different ingredients
Training for a marathon by running regularly
Which three components are at the core of any machine learning system?
Prediction, Coding, Testing
Representation, Evaluation, Optimization
Features, Models, Data
Algorithms, Math, Statistics
What happens when a model is too complex and learns noise instead of signal?
Generalization
Underfitting
Overfitting
Optimization
Imagine you’re teaching a robot to make coffee. Which feature would be most useful?
The color of the coffee mug
The day of the week
The number of beans used
The robot’s favorite TV show
A production line model predicts defects perfectly on last month’s data but misses new defect types this month. What happened?
Overfitting
Good generalization
Feature engineering success
Bias reduction
In demand forecasting, which principle helps decide whether to collect more sales data or engineer better features like promotions and seasonality?
Bias–Variance tradeoff
Learning curves
Ensemble methods
Theoretical guarantees
An ISE builds a model to predict machine breakdowns. What’s the WORST feature they could use?
Horoscope sign of the machine
Machine age
Operator shift
Vibration levels
According to Domingos, what usually matters more than choosing the fanciest algorithm?
Ensemble methods
More computer power
Using bigger models
Feature engineering
Which situation shows high variance (overfitting)?
A neural network predicting every tiny fluctuation in weekly demand
A linear regression missing seasonal cycles
A model achieving steady performance across datasets
A simulation with simplified assumptions
Why are ensembles (like bagging or boosting) often powerful?
They use less data
They average out errors
They remove the need for features
They are easier to interpret
Industrial & Systems Engineers can use ML for all of these EXCEPT:
Optimizing supply chain scheduling
Predicting patient flow in hospitals
Detecting defects with computer vision
Playing professional basketball
Biggest lesson from Domingos’ article?
Success in ML depends heavily on representation, evaluation, and features
Always choose the most advanced algorithm
Theory is always enough to guide practice
Data quality doesn’t matter as much
