I'm Klas and I'm a PhD student in the Accountable Systems Lab at Carnegie Mellon University, advised by Matt Fredrikson. My area of research is AI, and particularly, my research concentrates on understanding and demystifying deep learning. I work to improve the security, transparency, and generality of deep neural networks, with a focus on applications in computer vision and data privacy.

GHC 7004 | kleino cs. cmu. edu | klasleino

Research Interests

My area of research is AI, and particularly, my research concentrates on understanding and demystifying deep learning. I work to improve the security, transparency, and generality of deep neural networks, with a focus on applications in computer vision and data privacy. My work fits primarily under the sub-fields of explainable AI and ML security. Explainable AI aims to bring interpretability and transparency to otherwise opaque deep learning methods, giving us a richer understanding of their inner workings. ML security addresses concerns including attacks that compromise data privacy and that fool even state-of-the-art models. Currently, I am most interested in topics with three major themes; namely, explaining black-box neural network behavior, creating a theory of network generalization, and designing generative models for high-level feature interpretation and counterfactual reasoning.

Explaining Black-box Neural Network Behavior

In the recent years, deep neural networks have become increasingly powerful at tasks previously only humans had mastered. Deep learning has become widely used, and while it has many practitioners, its inner workings are far from well-understood. As the application of ML has increased, so has the need for algorithmic transparency, the ability to understand why algorithms deployed in the real world make the decisions they do. Much of my work has addressed the problem of determining which aspects of a network influence particular decisions, in addition to interpreting the identified influential components. Influence can be used to increase model trust, to uncover insights discovered by ML models, and as a building block for debugging arbitrary network behavior.

A Theory of Network Generalization

Despite having the capacity to significantly overfit, or moreover, memorize the training data, deep neural networks demonstrate an ability to generalize reasonably well in practice. Present hypotheses have failed to explain why this is the case. In fact, it is not well understood how exactly overfitting is manifested in a model. One aspect of my work tries to understand what phenomena give rise to misclassifications, overfitting, and bias in DNNs. Understanding the causes for these problems will also shed light on what leads models to generalize; and may suggest ways of improving generalization. Furthermore, as overfitting presents a threat to the security of a model, understanding overfitting more fundamentally may help protect the privacy of the data involved in training a model, and improve the model's robustness to adversarial manipulation. I develop explanations for these problems that have direct applications to membership inference, misclassification prediction, and bias amplification.

Generative Models

It is often difficult to interpret the high-level concepts learned by neural networks. I am interested in the use of generative models to explore the semantic space of deep networks. This would enable interpreting features that cannot be easily understood via their existence in a single instance. Furthermore, it may allow automated, interpretable counterfactual reasoning for DNNs.


Feature-wise Bias Amplification [ICLR 2019]

We study the phenomenon of bias amplification in classifiers, wherein a machine learning model learns to predict classes with a greater disparity than the underlying ground truth. We demonstrate that bias amplification can arise via an inductive bias in gradient descent methods that results in the overestimation of the importance of moderately-predictive "weak" features if insufficient training data is available. This overestimation gives rise to feature-wise bias amplification — a previously unreported form of bias that can be traced back to the features of a trained model. Through analysis and experiments, we show that while some bias cannot be mitigated without sacrificing accuracy, feature-wise bias amplification can be mitigated through targeted feature selection. We present two new feature selection algorithms for mitigating bias amplification in linear models, and show how they can be adapted to convolutional neural networks efficiently. Our experiments on synthetic and real data demonstrate that these algorithms consistently lead to reduced bias without harming accuracy, in some cases eliminating predictive bias altogether while providing modest gains in accuracy.

Influence-directed Explanations for Convolutional Neural Networks [ITC 2018]

We study the problem of explaining a rich class of behavioral properties of deep neural networks. Distinctively, our influence-directed explanations approach this problem by peering inside the network to identify neurons with high influence on a quantity and distribution of interest, using an axiomatically-justified influence measure, and then providing an interpretation for the concepts these neurons represent. We evaluate our approach by demonstrating a number of its unique capabilities on convolutional neural networks trained on ImageNet. Our evaluation demonstrates that influence-directed explanations (1) identify influential concepts that generalize across instances, (2) can be used to extract the "essence" of what the network learned about a class, and (3) isolate individual features the network uses to make decisions and distinguish related classes.


18-739: Security and Fairness of Deep Learning (Spring 2019)

This course will provide an introduction to deep learning methods with emphasis on understanding and improving their security, privacy, and fairness properties. The course will cover basics of machine learning and introduce popular deep learning methods. It will delve into applications of deep learning methods in security, their susceptibility to adversarial manipulation, and techniques for making deep learning robust to adversarial manipulation. It will cover state-of-the-art methods for explaining black-box deep learning models to enhance their transparency. It will also examine methods for deep learning that are designed to respect individual privacy and fairness. Students will do homework assignments and critique weekly readings. Prior knowledge of machine learning, deep learning, and security concepts are useful but not required.


15-781: Graduate Artificial Intelligence (Fall 2016)

15-122: Principles of Imperative Computation (Fall 2013)

15-122: Principles of Imperative Computation (Fall 2012)

  • Carnegie Mellon University

  • School of Computer Science

  • Computer Science Department

  • Accountable Systems Lab