fbpx
  • Start Time:
    03:00PM
  • End Time:
    03:20PM
  • Day:
    Day 2

Talk:

  • One of the key challenges with creating a supervised machine learning system is the availability of labelled data. Scenarios where labelled data is not available, we can use human domain knowledge along with Active Learning to create labelled data and use that to train a ML model
  •  In this talk, we can see how Active Learning was used within ING to create initial models for early warning detection. We used active learning to create a labelled dataset using a combination of sampling methods
  • The final models thus created were able to meet the KPI’s and are used in production environment as well

Associated Speakers:

Anil Panda

Data Scientist & ML Engineer - Risk & Pricing

ING

Associated Talks:

03:00PM - Day 2

View Solving a cold start problem in NLP with Active Learning

View Full Info