We thank Jim Martin, anonymous reviewers, and members of the NLP+CSS research group at CU Boulder for their insightful comments and discussions. This work was supported in part by NSF grant IIS-1849931.

In this work, we have considered a general framework for learning causal DAGs, which consists of PNS, structure learning, and pruning. We empirically study existing PNS methods and find that these methods may bring a negative effect on variable sele…

The work was jointly funded by the Ontario Ministry of Science and Innovation Early Researcher Award Number ER17-13-043; the Division of Infectious Diseases, St. Michael’s Hospital, University of Toronto; and the University of Toronto Work Study Pro…

Finding a good learning rate schedule is an old but crucially important issue in machine learning. This paper makes a step forward, proposing an automatic method to obtain performing LR schedules that uses an adaptive moving average over increasingl…

We have presented a local planner that is able to make kinematically feasible paths across an unknown off-road environment. We combined the classic A* algorithm along with a kinematic forward trajectory simulation to generate good enough paths. Our …

It has previously been demonstrated that CNNs are biased towards learning texture-based features over the shape-based features used in human vision. Although this bias aids classification performance when training and testing classes are drawn from …

This paper proposed a new optimization program to learn a time-varying Ising model with piecewise constant structure. To the best of our knowledge, it is the first to provide a change-point consistency theorem and a empirical comparative study with …

In this paper, we have presented a method of adaptively partitioning persistence diagrams for featurization using template functions. This allows for significant flexibility in the featurization for the user. Our methods are applicable to tent funct…

In this paper, we experimentally showed that end-to-end approaches and different orthographic units were rather suitable to model the French language. RNN-transducer was found specially competitive with character units compared to other end-to-end a…

This paper presents a novel concept pointer generator model to improve the abstractive summarization model and generate concept-oriented summaries. We also propose a novel distant supervision strategy for model adaption to different datasets. Both e…

We have introduced the framework of extremal perturbation analysis, which avoids some of the issues of prior work that use perturbations to analyse neural networks. We have also presented a few technical contributions to compute such extremal pertur…

Combining historical data from a large EMR database and state-of-the-art ML algorithms resulted in an average TAE benefit of 24.11%% (1.1 years) for patients diagnosed with CAD. Our results show that differing medication regimens and revascularizati…

Using an analogy between the gradient and the state-action value function, we have shown a strong connection between some constrained convex optimization and RL algorithms. We have made this connection clear on three cases: Frank Wolfe and Conservat…

We would like to thank Akshay Krishnamurthy, Chicheng Zhang, Mikael Henaff, and members of the Laboratory for Intelligent Probabilistic Systems for valuable discussion and feedback. JTA was supported by a Siemens FutureMakers graduate fellowship whi…

In this paper, we have presented a fast and easy method to synthesise training samples for the implementation of illumination-invariant models. The synthetic images are generated by artificially altering the pixel intensity values of the input image…

In this paper, we introduced a novel optimization technique, FOP, that learns a preconditioning matrix online to improve convergence and generalization in large-scale machine learning models. We tested FOP on several problems and architectures, exam…

We considered the problem of learning to explore autonomously in a non-stationary environment and proposed a pertinent performance measure. We gave a natural algorithm for the considered problem and proved an upper bound on the performance measure t…

In this paper, we propose a novel centroid updating approach to enhance the clustering-based superpixel methods and a superpixel based edge detection method. Experiments illustrate that our proposed methods can get a much better performance in noisy…

In this paper, we proposed a novel online-learning algorithm that generates new contextual identifier to aid learning of classification tasks under context or concept drifts. The building blocks of the algorithm consist of a task-oriented classifier…

Many open-domain NLP tasks rely upon multi-document input from the web to facilitate tasks such as answering questions or writing summaries, but current approaches struggle to encode the entirely of this information. We propose constructing one know…