University of Cambridge > Talks.cam > Engineering Safe AI > Decision Boundary Geometries and Robustness of Neural Networks

Log in

University Account

External (via Google)

Information on

Subscribing to talks details

Finding a talk details

Adding a talk details

Disseminating talks details

Help and Documentation details

Decision Boundary Geometries and Robustness of Neural Networks

Download to your calendar using vCal

Sven Wang (University of Cambridge)
Wednesday 14 February 2018, 17:00-18:30
Cambridge University Engineering Department, CBL Seminar room BE4-38. For directions see http://learning.eng.cam.ac.uk/Public/Directions.

If you have a question about this talk, please contact Adrià Garriga Alonso .

Adversarial examples are small perturbations to an input point that cause a Neural Network (NN) to misclassify it.

Some recent research shows the existence of “universal adversarial perturbations” which, unlike previous adversarial examples, are not specific to data points and network architectures. We will also talk about some results which try to link this behaviour to the geometry of decision boundaries learned by neural networks.

Adversarial inputs by themselves aren’t the main concern for the value alignment problem. However, the insight they can give about NN internals will be important if future AIs rely on NNs at all.

Relevant readings: The Robustness of Deep Networks: A Geometrical Perspective http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8103145&tag=1

Adversarial Spheres https://arxiv.org/abs/1801.02774

This talk is part of the Engineering Safe AI series.

This talk is included in these lists:

Note that ex-directory lists are not shown.

Decision Boundary Geometries and Robustness of Neural Networks

📅 Download to calendar (vCal)

👤 Speaker: Sven Wang (University of Cambridge)
📅 Date & Time: Wednesday 14 February 2018, 17:00 - 18:30
📍 Venue: Cambridge University Engineering Department, CBL Seminar room BE4-38. For directions see http://learning.eng.cam.ac.uk/Public/Directions

Questions? Contact Adrià Garriga Alonso

Abstract

Adversarial examples are small perturbations to an input point that cause a Neural Network (NN) to misclassify it.

Adversarial inputs by themselves aren’t the main concern for the value alignment problem. However, the insight they can give about NN internals will be important if future AIs rely on NNs at all.

Relevant readings: The Robustness of Deep Networks: A Geometrical Perspective http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8103145&tag=1

Adversarial Spheres https://arxiv.org/abs/1801.02774

Series This talk is part of the Engineering Safe AI series.

Included in Lists

Note: Ex-directory lists are not shown.

Log in

🔐 Log In

Information on

ℹ️ Information

Decision Boundary Geometries and Robustness of Neural Networks

This talk is included in these lists:

Decision Boundary Geometries and Robustness of Neural Networks

Abstract

Included in Lists

Log in

🔐 Log In

Information on

ℹ️ Information

Decision Boundary Geometries and Robustness of Neural Networks

This talk is included in these lists:

Other lists

Other talks

Decision Boundary Geometries and Robustness of Neural Networks

Abstract

Included in Lists