Three Short Talks from Prof. Daniel Weller's Group

  • Time: 1:00PM
  • Location: Rice 242

Title: How does the mouse brain work for spatial recognition?
Speaker: Haoyi Liang

Abstract: Numerous computer vision algorithms are proposed to teach computers to understand their environments, and artificial intelligence approaches, such as convolutional neural networks, show promising results. However, do animals perceive their environment in a similar way as AI? Our work with the Department of Neuroscience sheds light on this question by studying how mice perceive their environment.

Title: Smarter tuning of image processing algorithms through machine learning
Speaker: Tanjin Toma

Abstract: In image and video processing, algorithms for inverse problems (e.g., image enhancement, image reconstruction) often have some parameters which need to be set in order to yield good results. In practice, usually the choice of such parameters is made empirically with trial and error. But, manual tuning of parameters is time-consuming as well as impractical when multiple parameters exist in a problem. In this talk, I’ll discuss how machine learning can be exploited to automatically choose such parameters effectively.

Title: Robust myocardial T1 mapping with convolutional neural networks
Speaker: Haris Jeelani

Abstract: In cardiac magnetic resonance imaging, the T1 relaxation time in myocardial tissue may be used as an indicator for a variety of pathological conditions. A pixel-wise non-linear regression model is typically used to obtain T1 maps. In this talk we discuss our approach of obtaining T1 maps using convolutional neural networks that are more robust than the conventional regression method.

How to Train a More Interpretable Neural Text Classifier?

Hanjie Chen
Department of Computer Science
University of Virginia

  • Date: Wednesday April 17th, 2019
  • Time: 1:00PM
  • Location: Rice 242

Abstract Although neural networks have achieved remarkable performance on text classification, the lack of transparency causes the challenge of understanding model predictions. In the meantime, the growing demand for using neural networks in many text classification tasks drives the research of building more interpretable models.

In our work, we propose a novel training strategy, called learning with auxiliary examples, to improve the interpretability of existing neural text classifiers. By using sentiment classification as the example task and a well-adopted baseline convolutional neural network model as the neural classifier, we show that the new learning strategy improves the model interpretability while maintains similar classification performance. Besides, we also propose an automatic evaluation measurement to quantify the interpretability by measuring the consistency between the model predictions and the corresponding explanations. Experiments on two benchmark datasets show some significant improvements on the interpretability of the models trained with the proposed strategy.

Hypothesis Generation From Evolving Scientific Corpora - Promises and Challenges

Kishlay Jha
Department of Computer Science
University of Virginia

  • Date: Wednesday April 10th, 2019
  • Time: 1:00PM
  • Location: Rice 242

Abstract Hypothesis generation is a crucial element in making scientific discoveries. Traditionally, scientists form hypotheses based on their intuition, ability to make creative chance connections, prior knowledge and experience. This involves them to selectively read hundreds (sometimes thousands) of articles to develop testable hypotheses. However, in the present data-intensive era, it is infeasible for an individual scientist or a research team to keep up with all the relevant articles published in their area of interest. While technologies based on text summarization would help users get a high level idea of the papers, it fails to stitch together disparate and temporally evolving facts together to present novel and actionable insights that can drive new research frontiers.

To overcome these aforementioned challenges, in our research group, we have been developing a temporally robust computational framework that aims to identify implicit connections between hitherto unknown medical concepts, by modeling their semantic evolution over time. In this talk, I will present the “promises and challenges” of our initial steps taken towards this direction.

Improving Robustness of Neural Networks using Domain Knowledge

Weilin Xu
Department of Computer Science
University of Virginia

  • Date: Wednesday April 3rd, 2019
  • Time: 1:00PM
  • Location: Rice 242

Abstract Although machine learning techniques have achieved great success in many areas, such as computer vision, natural language processing, and computer security, recent studies have shown that they are not robust under attack. A motivated adversary is often able to craft input samples that force a machine learning model to produce incorrect predictions, even if the target model achieves high accuracy on normal test inputs. This raises great concern when machine learning models are deployed for security-sensitive tasks.

Our work aims to improve the robustness of machine learning models by exploiting domain knowledge. While domain knowledge has often been neglected due to the power of automatic representation learning in the deep learning era, we find that domain knowledge goes beyond a given dataset of a task and helps to (1) uncover weaknesses of machine learning models, (2) detect adversarial examples and (3) improve the robustness of machine learning models.

First, we design an evolutionary algorithm-based framework, Genetic Evasion, to find evasive samples. We embed domain knowledge into the mutation operator and the fitness function of the framework and achieve 100% success rate in evading two state-of-the-art PDF malware classifiers. Unlike previous methods, our technique uses genetic programming to directly generates evasive samples in the problem space instead of the feature space, making it a practical attack that breaks the trust of black-box machine learning models in the security field.

Second, we design an ensemble framework, Feature Squeezing, to detect adversarial examples against deep neural network models using simple pre-processing. We employ domain knowledge on signal processing that natural signals are often redundant for many perception tasks. Therefore, we can squeeze the input features to reduce adversaries’ search space while preserving the accuracy on normal inputs. We use various squeezers to pre-process an input example before it is fed into a model. The difference between those predictions is often small for normal inputs due to redundancy, while the difference can be large for adversarial examples. We demonstrate that Feature Squeezing is empirically effective and inexpensive in detecting adversarial examples generated by many algorithms.

Third, we incorporate simple pre-processing with certifiable robust training and formal verification to train provably-robust models. We formally analyze the implication of pre-processing on adversarial strength and derive a novel method to improve model robustness. We find that our approach produces accurate models with verified state-of-the-art robustness and advances the state-of-the-art of certifiable robust training methods.

We demonstrate that domain knowledge helps us understand and improve the robustness of machine learning models. Our results have motivated several subsequent works, and we hope this talk will be inspirational to implement robust models under attack.

Insights from Social Psychology to Computational User Modeling

Lin Gong
Department of Computer Science
University of Virginia

  • Date: Wednesday March 27th, 2019
  • Time: 1:00PM
  • Location: Rice 242

Abstract The advent of participatory web has created massive amounts of user-generated data, which enables the study of online user attributes and behaviors. Traditional social psychology studies commonly conduct surveys and experiments to collect user data in order to infer attributes of individuals, which are expensive and time-consuming. In contrast, we aim to understand users by building computational user models automatically, thereby to save time and efforts. And the principles of social psychology serve as good references for building such computational models.

In this presentation, I will discuss about my PhD works for modeling online user behaviors based on user-generated data. In particular, two challenges are addressed: (1) modeling users’ diverse ways of expressing attitudes or opinions; (2) building unified user models by integration of different modalities of user-generated data. This presentation bridges the gap between social psychology and computation behavior modeling which also provides a foundation for making user behavior modeling useful for many other applications.