2. ● Exploratory Data Analysis (EDA)
● Training Model
● Deploy Model
● Monitor model over time (maintenance)
● Scale model as it gets more usage
Enterprise Deep Learning workflows
3. ● Infrastructure for USING Deep Learning
● “Serving” models to end users
● Visualization
● Auditing of data flow (Where did that come from?)
● Bundled hardware acceleration
Training
4. ● Need to visualize
● Neural nets aren’t interpretable
● DL has its own vocabulary in addition to “Machine learning”
● Hard to track research from practical
● Not much emphasis on “apps”
Why is training “hard”?
9. ● Infrastructure for USING Deep Learning
● “Serving” models to end users
● Visualization
● Auditing of data flow (Where did that come from?)
● Bundled hardware acceleration
Deployment