Algorithms
to Live By
Better Summary
Owlet Library
Ferdinand Chandra
Optimal Stopping
37% rule
Look then leap
Gather info
Then commit
Secretary problem
Candidates sequential
Reject gone
Binary decision
Yes hire
No reject
10 candidates
37% is 4
Look phase
No hire
Learn good
Leap phase
Hire better
50% optimal
Dating tweaks
Rejection possible
Can return
Full information
Selling property
Know costs
Maximize profit
Wait costs
Look then leap
Stop crucial
Explore Exploit
Time constraints
Plenty time
Explore options
Short time
Exploit known
Optimism prevents regret
Explore fail
Better than regret
Sort
Nothing interesting
Cache
Forgetting curve
Brain caches
Relevant info
Cache shows importance
LRU algorithm
Linear time
Effective indexing
Bookshelf example
Keep close
Rarely used
Into storage
Scheduling
Minimize lateness
Earliest due first
Minimize completion times
Shortest processing first
Homework easiest subjects
Context switching overhead
Thrashing
No real work
Start doing
Work dumber
Interrupt coalescing
Batch recurring tasks
No perfect schedule
Optimize for you
Bayes Rule
Laplace’s Law
Unknown tickets
Draw trials
n draws
w wins
Next win
(w+1)/(n+2)
Sunrise example
5000 years
Still not 100
Kind reply example
(2+1)/(2+2)
Three quarters
Copernican Principle
Wall stood long
Stand another period
Past data insufficient
Choose priors
Overfitting
Against complexity
Simple model better
More data hurts
Complexity traps noise
Simplicity sharpens
Partner checklist
One question better
Overfitting definition
Tailored data
Fails new
Cross-validation
Test unseen data
Exams analogy
Memorization vs knowledge
Penalize complexity
Lasso method
Elevator pitch
Proverbs wisdom
Think less
Uncertainty simple
Relaxation
Constraint Relaxation
Loosen restrictions
MST example
Real life stuck
Too many rules
Continuous Relaxation
Fractional values
Day planning
Gym 0.4
Work 0.5
Netflix 0.1
Round by priority
Lagrangian Relaxation
Bake constraints
Scoring system
Lunch budget
Taste minus price
Compare options
Randomness
Logic fails
Planning fails
Optimization fails
Randomness fallback
Avoid bias
Avoid deadlock
Miller-Rabin algorithm
Primality testing
Fast accurate
Hill Climbing
Escape local maxima
Break mental bubble
Try new things
Stimulate creativity
Network
Exponential Backoff
Wait doubles
Failed attempts
Prevent overload
Arguments back off
Job applications
Don’t spam
Wait longer
Better never late
Delayed responses judged
No reply missed
Late reply ignored
Buffered world
Network delays
YouTube buffers
Human relationships
Asynchronous communication
Game Theory
Recursion
Thinking about thinking
Level 0 random
Level 1 opponent
Level 2 nested
Level 3 nested
Same level needed
Overcomplicate mistakes
Poker bluff
Backfire
Equilibrium
No player improve
Two-player games
At least one
Rock-Paper-Scissors
Random equal
Stable equilibrium
Tragedy Commons
Shared resource
Self-interest overuse
Depletion
Must-take rule
Equilibrium no take
Anticipate actions
Underuse
Mechanism Design
Reverse game theory
Design rules
Desired outcomes
Hate game
Fix incentives
Information Cascades
Bad collective outcomes
Bubbles
Dutch auction
Price drops
Recursive reasoning
System collapse
Honest valuation best
Tragic Rationality
Individual gain
Collective inefficiency
Conclusion
Better summary
Owlet Library
Ferdinand Chandra
Book notes
Computer science
Human decisions