Travis Green, Max Lin, Robert Kaplow, Jóhannes Kristinsson, Ryan McGee
What is a smart app?
1. Automates the repetitive
2. Recommends the useful
3. Extracts the essential
4. Pandora, Cabsense, etc.
What is machine learning?
A set of algorithms that learn patterns from data and make intelligent decisions
input -> predictive model -> output
How do I build one?
Train the model
Building a smart app
Step 1: Upload your training data
– training data: output and input features
– data format: CSV
Step 2: Train
Create a new model by training on data
Step 3: Prediect
Apply the trained model to make predictions on new data
Connectivity with Sync and the cloud
Personalized driving experience
Plug in hybrid to identify zones where to save energy
Using driving history, predict optimization, optimize powertrain
Collect and analyze the date
Empowering the car using the cloud
How can I add more data?
Streaming training to add real-time data to your predictive model
New API feature
– adapt quickly to new data
– automatically improve performance over time
– alternate way to train predictive models
Step 4: Adapt
Steam new data to your predictive model
What if I need data?
Hosted model subscriptions
– Users can subscribe to others' models
– Hosted model revenue shared with model developer
– Advantages to users data already gathered and labeled, built by experts, easy to add to your app
Access any model just like a normal prediction
All models are already enable for all users
Available now: sentiment, tagger, languageid