Airstream use case Houssam Nassif AISTATS’15.

Recommendation challenges at Amazon

Airstream use case

Houssam NassifAISTATS’15

Mission Statement

To be Earth’s Most Customer-Centric Company Where People Can Find and Discover Anything

They Want to Buy Online

Airstream

Visual Browse Experience

• How to define and measure?• Image parameters? Clicks? Hand curated?

Beautiful

• How to engage users?• Optimize for discovery or repeatability?• What products to show?

Engaging

Optimize for Users

• 244MM active accounts• 14 countries

Customers

• Personalized• What metric to optimize?• Milliseconds latency

Recommendation at scale

Temporal Variations

Time (in hours)

Val

ue (

clic

ks,

purc

hase

s)

Zero-Inflated DistributionC

usto

mer

s

Engagement

How to Recommend?

Multi-armed bandits

Collaborative filtering

Deep learning

Regression analysis …

Need for Diversity

How to Diversify?

Challenges• How to learn ideal mix?• How to balance between diversity and

metric of interest?

Possible solutions• Determinantal Point Processes • Submodular Functions

ML Teams @ Amazon

ML SeattleML Berlin

ML Bangalore

S9

A9

A2Z

Toronto Dev Center

ML Bay Area