Top Banner
PROOF OF FAILURE Clare Corthell Machine Learning Engineer & Data Scientist @clarecorthell www.datasciencemasters.org
18

User-Operated Model-Building Systems - Data Science: Inconvenient Truths

Aug 05, 2015

Download

Technology

Clare Corthell
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

PROOF OF FAILURE

Clare CorthellMachine Learning Engineer & Data Scientist

@clarecorthellwww.datasciencemasters.org

Page 2: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

Deal Intelligence Platformfind and evaluate private companies

Page 3: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

Machine Learning Need

• Very little structured information

• Disaggregated data

• Need for categorization

=> Data Structuring & Creation

Page 4: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

WHAT DOES THE COMPANY DO?“industry” dimension

Page 5: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

INDUSTRIES AS BINARY CATEGORIESyou’re in or out

Page 6: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

inputs outputsmodel

decision:• reinforce• ship

Page 7: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

PERFECTION! UTOPIA!

HUMAN INFERENCE WITHOUT HUMANS!and 60 fewer people on payroll

- $4.2m / yr

Page 8: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

ANALYSTSthe user is not the database

Page 9: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

EXAMPLE 1: WIND TURBINESWind Turbines.

definition

Page 10: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

EXAMPLE 2: WEARABLES

on your body?electronic?

new materials?what are they?

definition

Page 11: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

REINFORCEMENTdoesn’t always work

Page 12: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

REINFORCEMENT PITFALLS

- (Technical) Overfitting- Humans have to question their own assumptions- Dimensional encoding issues (is this expressible in features?)- Human definitions is inadequate

Page 13: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

SOOOOOOOOOOOOOOO…

SVM > neural nets

things I’ve heard recently

Page 14: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

WHY IS SVM BETTER?Feature inspectability

• sometimes for debugging• mostly for humans

Humans don’t know what transformation the black box exerts on inputs. But sometimes, they need to know.

Their investors, their customers, their data analysts, their operators, their CEO — all want to know.

Page 15: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

MONSTERS IN THE BLACK BOX

because

Page 16: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

HUMANS SHOULD BE HUMANSCOMPUTERS SHOULD BE COMPUTERS.

Sometimes, our identities get a little mixed up.

Page 17: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

1. Set Expectationsmake sure the organization understands failures

2. Reduce the “Trickery”*We build systems for humans. They need to understand how the levers and knobs affect the outcome

*h/t Sean Taylor

Page 18: User-Operated Model-Building Systems - Data Science: Inconvenient Truths

datasciencemasters.org

[email protected]

@clarecorthell

mattermark.com