1 Introduction to Human-in-the-Loop Machine Learning

Unlike robots in the movies, most of today’s Artificial Intelligence (AI) cannot learn by itself: it relies on intensive human feedback. Probably 90% of Machine Learning applications today are powered by Supervised Machine Learning. This covers a wide range of use cases: an autonomous vehicle can drive you safely down the street because humans have spent thousands of hours telling it when its sensors are seeing a ‘pedestrian’, ‘moving vehicle’, ‘lane marking’, and every other relevant object; your in-home device knows what to do when you say ‘turn up the volume’, because humans have spent thousands of hours telling it how to interpret different commands; and your Machine Translation service can translate between languages because it has been trained on thousands (or maybe millions) of human-translated texts.

1.1       The Basic Principles of Human-in-the-Loop Machine Learning

1.2       Introducing Annotation

1.2.1   Simple and more complicated annotation strategies

1.2.2   Plugging the gap in data science knowledge

1.2.3   Quality human annotations: why is it hard?

1.3       Introducing Active Learning: improving the speed and reducing the cost of training data

1.3.1   Three broad Active Learning sampling strategies: uncertainty, diversity, and random

1.3.2   What is a random selection of evaluation data?

1.3.3   When to use Active Learning?

1.4       Machine Learning and Human-Computer Interaction

1.4.1   User interfaces: how do you create training data?

1.4.2   Priming: what can influence human perception?

1.4.3   The pros and cons of creating labels by evaluating Machine Learning predictions

1.4.4   Basic principles for designing annotation interfaces

1.5       Machine Learning-Assisted Humans vs Human-Assisted Machine Learning

1.6       Transfer learning to kick-start your models

1.6.1   Transfer Learning in Computer Vision

1.6.2   Transfer Learning in Natural Language Processing

1.7       What to expect in this text

1.8       Summary