Loading...

Andreas Dewes - Fairness and transparency in machine learning: Tools and techniques

374 views

Loading...

Loading...

Transcript

The interactive transcript could not be loaded.

Loading...

Rating is available when the video has been rented.
This feature is not available right now. Please try again later.
Published on Jul 26, 2017

Description
This talk will try to answer a simple question: When building machine learning systems, how can we make sure that they treat people fairly and can be held accountable? While seemingly trivial, this question is not easy to answer, especially when using complex methods like deep learning. I will discuss tools and techniques that we can use to make sure our algorithms behave as they should.

Abstract
When working with personal data, we need to make sure that our algorithms treat people fairly, are transparent and can be held accountable for their decisions. When using complex techniques like deep learning on very large datasets, it is not easy to prove that our algorithms behave they way we intend them to and e.g. do not discriminate against certain groups of people.

In my talk, I will discuss why ensuring transparency and fairness in machine learning is not easy, and how we can use Python tools to investigate our machine learning systems and makre sure they behave they way they should.

-Introduction: Why you should care about this (EU Data Protection Directive)
-What kind of problems can occur in machine learning systems (bias in the input data, leakage of sensitive information into the training data, hidden usage of protected attributes by the algorithm)?
-How can we measure and correct for bias in our systems (certifying and removing disparate impact)?
-How can we understand the decisions that our algorithms make (perturbation analysis, simplified modeling, blackbox testing)?
-How can we design our machine learning systems to make sure they're compliant and accountable (anonymization of data, monitoring of outcomes, auditing of algorithms)?
-Outlook: The future of transparency and accountability in machine learning

www.pydata.org

PyData is an educational program of NumFOCUS, a 501(c)3 non-profit organization in the United States. PyData provides a forum for the international community of users and developers of data analysis tools to share ideas and learn from each other. The global PyData network promotes discussion of best practices, new approaches, and emerging technologies for data management, processing, analytics, and visualization. PyData communities approach data science using many languages, including (but not limited to) Python, Julia, and R.

PyData conferences aim to be accessible and community-driven, with novice to advanced level presentations. PyData tutorials and talks bring attendees the latest project features along with cutting-edge use cases.

Comments are disabled for this video.

to add this to Watch Later

Add to

Loading playlists...