Machine learning isn’t a set-it-and-forget-it operation. Even with solid examples, ML algorithms can still fail and end up blocking important emails, filtering out useful content, and causing a variety of other problems. In this O’Reilly report, industry analyst Ted Cuzzillo examines real-world examples of active learning, a relatively new strategy for improving ML results through short-term human intervention.
Throughout this report, Cuzzillo relies on several experts in the field for practical applications and tips they’ve unearthed through various projects in active learning. As you’ll discover, the point at which algorithms fail is precisely where there’s an opportunity to insert human judgment to actively improve the algorithm’s performance.
- Learn the basic principles of labeling data
- Help algorithms decrease the uncertainty of their results in email spam filtering and online search accuracy
- Effectively use crowdsourcing in your ML training project, without incurring undue costs
- Select training data from areas where the data volume is greatest
- Use multiple classification methods to better train ML algorithms
- Learn how to apply more advanced strategies for even greater accuracy