Transferring Deep Learning into a Recommender System By Christophe - - PowerPoint PPT Presentation

Transferring Deep Learning into a Recommender System

By Christophe Duong, Data Scientist

Big Data Paris 2017

• High-end Product
• > Appearance is crucial
• Fewer recurrent buyers
• > Web visit patterns are essential
• > Short visit sessions (browsing 4-10 first

sales)

• Challenging context for standard

recommender systems

• Mostly flash sales, i.e. volatile, ephemeral
• > Unlike amazon / Price Minister / Cdiscount
• > Sales are seasonal (Christmas, ski,

summer)

A data science workflow with

Six steps to a predictive model

Data Exploration & Understandin g Data Preparation Model Creation Evaluation Deployment Data Acquisition

Dataset 1

Scored dataset Scored dataset

Iteration 1 Iteration 2 Iteration n

Creating a predictive model is a highly iterative process. Data Science Studio enables its users to create and manage these projects from end-to-end. This process is not industry specific, and can be applied to many use cases.

Dataset 2 Dataset n

Business Understandin g

Adapted from the CRISP-DM methodology

Basic Recommendation Engines

Other Factors

One Meta Model to Rule Them All

Recommenders as features Machine learning to optimize purchasing probability Combine Recommend Describe

Cleaning, combining and enrichment of data Recommendation Engines Optimization of home display

the application automatically runs and compiles heterogeneous data Generation of recommendations based on user behaviour

Every customer is shown the 10 sales he is the most likely to buy Customer visits Purchases

Meta model combine recommendations to directly optimize purchasing probability

Meta Model

Recommender system for Home Page Ordering

+7% revenue

Sales information Sales Images

(A/B testing)

Batch Scoring every night

Sales Images

Integrating Image Information

Labelling Model

Pool + Palm Trees Hotel + Mountains Pool + Forest + Hotel + Sea Sea + Beach +Forest + Hotel Sales descriptions

CONTENT BASED

Recommender System

Using Deep Learning models

Common Issues

“I don’t have GPUs server” “I don’t have a deep leaning expert” “I don’t have labelled data” (or too few) “I don’t have the time to wait for model training ” I don’t want to pay for private apis” / “I’m afraid their labelling will change over time”

Solution 1 : Pre trained models

“I don’t have (or few) labelled data”

• > Is there similar data ?

PLACES DATABASE VPG SUN DATABASE 205 categories 2.5 M images 307 categories 110 K images

tower: 0.53 skyscraper: 0.26 swimming_pool/outdoor: 0.65 inn/outdoor: 0.06

Solution 1 : Pre trained models

If there is open data, there is an open pre trained model !

• Kudos to the community
• Check the licensing

Example with Places (Caffe Model Zoo) :

Solution 2 : Transfer Learning

“I want to add information of SUN database” “But I have only 100 K images”

If you know how to recognize… after a little bit of training… you will be able to recognize Transfer Learning

Use a network that knows how to see

• As a feature generator / transformer
• To be updated for the new problem

PLACES DATABASE VOYAGE PRIVE SUN DATABASE Training (optional) Pre-trained model VGG16 tower: 0.53 skyscraper: 0.26 Re-Training Transferred Data : Last convolutional layer features Re-trained model TensorFlow 2 fully connected layers Caffe Model Zoo GPU CPU GPU

Leverage existing knowledge !

Solution 2 : Transfer Learning

Accuracy: 72%, Top-5 Acc: 90 % > state of the art on dataset alone

Resulting Labels

Complementary information

Redondant information

Issue with our approach:

Solution : Matrix Factorization = 200x200 pixels -> 600 tags => 30 themes

Image content detection

Topic scores determine the importance of topics in an image

TOPIC TOPIC SCORE (%) Golf course – Fairway – Putting green 31 Hotel – Inn – Apartment building outdoor 30 Swimming pool – Lido Deck – Hot tub outdoor 22 Beach – Coast - Harbor 17 TOPIC TOPIC SCORE (%) Tower – Skyscraper – Office building 62 Bridge – River – Viaduct 38

Results ?

All Visits :

• Mostly France
• Pool displayed

First Recommendation

• Fail to display pools

Only Images ?

• Pool all around the world

Third column = Mix

What’s Next ?

Kenya Pragu e Berlin Cambodia

Conclusion

• Deep Learning ?
• Is there existing data ?
• Is there a pre-trained model ?
• Transfer Learning
• Cheaper, faster
• Any Data Scientist can do it
• Do Agile Data Science

!

• Start simple
• Validate each steps
• Iterate and grow