Modeling Mindsets: The Many Cultures Of Learning From Data

Preface
Who This Book is For
Introduction
	Models Have Variables And Learnable Functions
	Models Are Embedded In Mindsets
	A Mindset Is A Perspective Of The World
	Mindsets Are Cultural
	Mindsets Are Archetypes
	Mindsets Covered In This Book
Statistical Modeling – Reason Under Uncertainty
	Every ``Thing\'\' Has A Distribution
	Models Encode The Data-Generating Process
	Good Models Satisfy Assumptions And Fit Data
	Models Enable Conclusions
	Strengths & Limitations
Frequentism – Infer ``True\'\' Parameters
	Probability Is A Long-Run Frequency
	Imagined Experiments Underpin Inference
	Decide With Tests And Intervals
	Strengths & Limitations
Bayesianism – Update Parameter Distributions
	Bayes Demands A Prior
	The Likelihood Unites All Statistical Mindsets
	The Posterior Is The Modeling Target
	Use The Posterior To Learn About The World
	Strengths & Limitations
Likelihoodism – Likelihood As Evidence
	Statistical Mindsets Use Likelihood Differently
	Get Rid Of Priors And Imagined Experiments
	Compare Hypotheses Using Likelihoods
	Strengths & Limitations
Causal Inference – Identify And Estimate Causes
	Causality Is Often Ignored
	Visualize Causality With DAGs
	Pick A Flavor Of Causality
	A Causal Model Comes Before Estimation
	Strengths & Limitations
Machine Learning – Learn Algorithms From Data
	Put The Computer First
	Focus On Task Performance
	Machine Learning As Statistical Learning
	Strengths & Limitations
Supervised Learning – Predict New Data
	Learning Is Optimization And Search
	Good Models Predict New Data Well
	Supervision Enables Automation And Competition
	Mimic Outputs, Not The Process
	Strengths & Limitations
Unsupervised Learning – Find Hidden Patterns
	Find Patterns In The Data Distribution
	Unsupervised Learning Has Many Tasks
	Strengths & Limitations
Reinforcement Learning – Learn To Interact
	The Model Acts In A Dynamic World
	Learn In Different Ways
	Combine Reinforcement With Deep Learning
	Strengths & Limitations
Deep Learning - Learn End-To-End Networks
	Mindset Emerges From Neural Networks
	Modularity Allows End-To-End Modeling
	Properties Emerge From Neural Networks
	Strengths & Limitations
The T-Shaped Modeler
	Pragmatic Modeling Requires Many Mindsets
	Don\'t Try To Be An Expert On All Mindsets
	Become A T-Shaped Modeler
Acknowledgments
References