ml-debug/docs/evidence/cs229_ml_advice.md

Source: https://cs229.stanford.edu/materials/ML-advice.pdf Title: CS229 - Advice for Applying Machine Learning (Andrew Ng) Fetched-via: bash -c 'uvx "markitdown[pdf]" https://cs229.stanford.edu/materials/ML-advice.pdf' Fetch-status: verbatim

Advice for applying Machine Learning

Andrew Ng

Stanford University

Andrew Y. Ng

Today’s Lecture

• Advice on how getting learning algorithms to different applications.

• Most of today’s material is not very mathematical. But it’s also some of the

hardest material in this class to understand.

• Some of what I’ll say today is debatable.

• Some of what I’ll say is not good advice for doing novel machine learning

research.

• Key ideas:

1. Diagnostics for debugging learning algorithms.
2. Error analyses and ablative analysis.
3. How to get started on a machine learning problem.

– Premature (statistical) optimization.

Andrew Y. Ng

Debugging Learning Algorithms

Andrew Y. Ng

Debugging learning algorithms

Motivating example:

• Anti-spam. You carefully choose a small set of 100 words to use as

features. (Instead of using all 50000+ words in English.)

• Bayesian logistic regression, implemented with gradient descent, gets 20%

test error, which is unacceptably high.

• What to do next?

Andrew Y. Ng

Fixing the learning algorithm

• Bayesian logistic regression:

• Common approach: Try improving the algorithm in different ways.

– Try getting more training examples. – Try a smaller set of features. – Try a larger set of features. – Try changing the features: Email header vs. email body features. – Run gradient descent for more iterations. – Try Newton’s method. – Use a different value for λ. – Try using an SVM.

• This approach might work, but it’s very time-consuming, and largely a matter

of luck whether you end up fixing what the problem really is.

Andrew Y. Ng

Diagnostic for bias vs. variance

Better approach:

– Run diagnostics to figure out what the problem is. – Fix whatever the problem is.

Bayesian logistic regression’s test error is 20% (unacceptably high).

Suppose you suspect the problem is either:

– Overfitting (high variance). – Too few features to classify spam (high bias).

Diagnostic:

– Variance: Training error will be much lower than test error. – Bias: Training error will also be high.

Andrew Y. Ng

More on bias vs. variance

Typical learning curve for high variance:

r o r r e

Test error

Desired performance

Training error

m (training set size)

• Test error still decreasing as m increases. Suggests larger training set will help. • Large gap between training and test error.

Andrew Y. Ng

More on bias vs. variance

Typical learning curve for high bias:

r o r r e

Test error

Training error

Desired performance

m (training set size)

• Even training error is unacceptably high. • Small gap between training and test error.

Andrew Y. Ng

Diagnostics tell you what to try next

Bayesian logistic regression, implemented with gradient descent.

Fixes to try:

– Try getting more training examples. – Try a smaller set of features. – Try a larger set of features. – Try email header features. – Run gradient descent for more iterations. – Try Newton’s method. – Use a different value for λ. – Try using an SVM.

Fixes high variance. Fixes high variance. Fixes high bias. Fixes high bias.

Andrew Y. Ng

Optimization algorithm diagnostics

• Bias vs. variance is one common diagnostic.

• For other problems, it’s usually up to your own ingenuity to construct your

own diagnostics to figure out what’s wrong.

• Another example:

– Bayesian logistic regression gets 2% error on spam, and 2% error on non-spam.

(Unacceptably high error on non-spam.)

– SVM using a linear kernel gets 10% error on spam, and 0.01% error on non-

spam. (Acceptable performance.)

– But you want to use logistic regression, because of computational efficiency, etc.

• What to do next?

Andrew Y. Ng

More diagnostics

• Other common questions:

– Is the algorithm (gradient descent for logistic regression) converging?

J(θ)

e v i t c e b O

j

Iterations

It’s often very hard to tell if an algorithm has converged yet by looking at the objective.

Andrew Y. Ng

More diagnostics

• Other common questions:

– Is the algorithm (gradient descent for logistic regression) converging? – Are you optimizing the right function? – I.e., what you care about:

(weights w(i) higher for non-spam than for spam). – Bayesian logistic regression? Correct value for λ?

– SVM? Correct value for C?

Andrew Y. Ng

Diagnostic

An SVM outperforms Bayesian logistic regression, but you really want to deploy Bayesian

logistic regression for your application.

Let θSVM be the parameters learned by an SVM.

Let θBLR be the parameters learned by Bayesian logistic regression.

You care about weighted accuracy:

θSVM outperforms θBLR. So:

BLR tries to maximize:

Diagnostic:

Andrew Y. Ng

Two cases

Case 1:

But BLR was trying to maximize J(θ). This means that θBLR fails to maximize J, and the

problem is with the convergence of the algorithm. Problem is with optimization algorithm.

Case 2:

This means that BLR succeeded at maximizing J(θ). But the SVM, which does worse on

J(θ), actually does better on weighted accuracy a(θ).

This means that J(θ) is the wrong function to be maximizing, if you care about a(θ).

Problem is with objective function of the maximization problem.

Andrew Y. Ng

Diagnostics tell you what to try next

Bayesian logistic regression, implemented with gradient descent.

Fixes to try:

– Try getting more training examples. – Try a smaller set of features. – Try a larger set of features. – Try email header features. – Run gradient descent for more iterations. – Try Newton’s method. – Use a different value for λ. – Try using an SVM.

Fixes high variance. Fixes high variance. Fixes high bias. Fixes high bias. Fixes optimization algorithm. Fixes optimization algorithm. Fixes optimization objective. Fixes optimization objective.

Andrew Y. Ng

The Stanford Autonomous Helicopter

Payload: 14 pounds Weight: 32 pounds

Andrew Y. Ng

Machine learning algorithm

1. Build a simulator of helicopter.

Simulator

2. Choose a cost function. Say J(θ) = ||x – xdesired||2 (x = helicopter position)

3. Run reinforcement learning (RL) algorithm to fly helicopter in simulation, so

as to try to minimize cost function:

θRL = arg minθ J(θ)

Suppose you do this, and the resulting controller parameters θRL gives much worse

performance than your human pilot. What to do next?

Improve simulator? Modify cost function J? Modify RL algorithm?

Andrew Y. Ng

Debugging an RL algorithm

The controller given by θRL performs poorly. Suppose that:

1. The helicopter simulator is accurate.

2. The RL algorithm correctly controls the helicopter (in simulation) so as to

minimize J(θ).

3. Minimizing J(θ) corresponds to correct autonomous flight.

Then: The learned parameters θRL should fly well on the actual helicopter.

Diagnostics:

If θRL flies well in simulation, but not in real life, then the problem is in the simulator. Otherwise:

2. Let θhuman be the human control policy. If J(θhuman) < J(θRL), then the problem is in the reinforcement learning algorithm. (Failing to minimize the cost function J.) If J(θhuman)

J(θRL), then the problem is in the cost function. (Maximizing it

≥

doesn’t correspond to good autonomous flight.)

Andrew Y. Ng

More on diagnostics

• Quite often, you’ll need to come up with your own diagnostics to figure out

what’s happening in an algorithm.

• Even if a learning algorithm is working well, you might also run diagnostics to

make sure you understand what’s going on. This is useful for:

– Understanding your application problem: If you’re working on one important ML

application for months/years, it’s very valuable for you personally to get a intuitive understand of what works and what doesn’t work in your problem.

– Writing research papers: Diagnostics and error analysis help convey insight about

the problem, and justify your research claims.

– I.e., Rather than saying “Here’s an algorithm that works,” it’s more interesting to say “Here’s an algorithm that works because of component X, and here’s my justification.”

• Good machine learning practice: Error analysis. Try to understand what

your sources of error are.

Andrew Y. Ng

Error Analysis

Andrew Y. Ng

Error analysis

Many applications combine many different learning components into a “pipeline.” E.g., Face recognition from images: [contrived example]

Camera image

Preprocess (remove background)

Eyes segmentation

Face detection

Nose segmentation

Logistic regression

Label

Mouth segmentation

Andrew Y. Ng

Camera image

Preprocess Preprocess (remove background) (remove background)

Error analysis

Eyes segmentation Eyes segmentation

Face detection Face detection

Nose segmentation Nose segmentation

Logistic regression Logistic regression

Label

Mouth segmentation Mouth segmentation

How much error is attributable to each of the

components?

Plug in ground-truth for each component, and

see how accuracy changes.

Conclusion: Most room for improvement in face

detection and eyes segmentation.

Component

Accuracy

Overall system

85%

Preprocess (remove background)

Face detection

Eyes segmentation

Nose segmentation

Mouth segmentation

85.1%

91%

95%

96%

97%

Logistic regression

100% Andrew Y. Ng

Ablative analysis

Error analysis tries to explain the difference between current performance and

perfect performance.

Ablative analysis tries to explain the difference between some baseline (much

poorer) performance and current performance.

E.g., Suppose that you’ve build a good anti-spam classifier by adding lots of

clever features to logistic regression:

– Spelling correction. – Sender host features. – Email header features. – Email text parser features. – Javascript parser. – Features from embedded images.

Question: How much did each of these components really help?

Andrew Y. Ng

Ablative analysis

Simple logistic regression without any clever features get 94% performance.

Just what accounts for your improvement from 94 to 99.9%?

Ablative analysis: Remove components from your system one at a time, to see

how it breaks.

Component

Accuracy

Overall system

Spelling correction

Sender host features

Email header features

Email text parser features

Javascript parser

Features from images

99.9%

99.0

98.9%

98.9%

95%

94.5%

94.0%

[baseline]

Conclusion: The email text parser features account for most of the

improvement.

Andrew Y. Ng

Getting started on a learning problem

Andrew Y. Ng

Getting started on a problem

Approach #1: Careful design.

• Spend a long term designing exactly the right features, collecting the right dataset,

and designing the right algorithmic architecture.

•

Implement it and hope it works.

• Benefit: Nicer, perhaps more scalable algorithms. May come up with new, elegant,

learning algorithms; contribute to basic research in machine learning.

Approach #2: Build-and-fix.

•

Implement something quick-and-dirty.

• Run error analyses and diagnostics to see what’s wrong with it, and fix its errors.

• Benefit: Will often get your application problem working more quickly. Faster time to

market.

Andrew Y. Ng

Premature statistical optimization

Very often, it’s not clear what parts of a system are easy or difficult to build, and

which parts you need to spend lots of time focusing on. E.g.,

Camera image

Preprocess (remove background)

This system’s much too complicated for a first attempt.

Eyes segmentation

Step 1 of designing a learning system: Plot the data.

Face detection

Nose segmentation

Logistic regression

Label

The only way to find out what needs work is to implement something quickly,

and find out what parts break.

Mouth segmentation

[But this may be bad advice if your goal is to come up with new machine

learning algorithms.]

Andrew Y. Ng

The danger of over-theorizing

3d similarity learning

Color invariance

Object detection

Navigation

Differential geometry of 3d manifolds

Complexity of non-Riemannian geometries

VC dimension

… Convergence

bounds for sampled non- monotonic logic

Mail delivery robot

Obstacle avoidance

Robot manipulation

[Based on Papadimitriou, 1995]

Andrew Y. Ng

Summary

Andrew Y. Ng

Summary

• Time spent coming up with diagnostics for learning algorithms is time well-

spent.

•

It’s often up to your own ingenuity to come up with right diagnostics.

• Error analyses and ablative analyses also give insight into the problem.

• Two approaches to applying learning algorithms:

– Design very carefully, then implement.

• Risk of premature (statistical) optimization. – Build a quick-and-dirty prototype, diagnose, and fix.

Andrew Y. Ng