With the advent of large language models (LLM), the line between human-crafted and machine-generated texts has become increasingly blurred.This paper delves into the inquiry of identifying discernible and unique linguistic properties in texts that were written by humans, particularly uncovering the underlying discourse structures of texts beyond their surface structures.Introducing a novel methodology, we leverage hierarchical parse trees and recursive hypergraphs to unveil distinctive discourse patterns in texts produced by both LLMs and humans.Empirical findings demonstrate that, although both LLMs and humans generate distinct discourse patterns influenced by specific domains, human-written texts exhibit more structural variability, reflecting the nuanced nature of human writing in different domains.Notably, incorporating hierarchical discourse features enhances binary classifiers' overall performance in distinguishing between human-written and machine-generated texts, even on out-of-distribution and paraphrased samples.This underscores the significance of incorporating hierarchical discourse features in the analysis of text patterns.The code and dataset are available at https://github.com/ minnesotanlp/threads-of-subtlety.

Class Activation Mapping (CAM) is a powerful technique used to understand the decision making of Convolutional Neural Network (CNN) in computer vision. Recently, there have been attempts not only to generate better visual explanations, but also to improve classification performance using visual explanations. However, previous works still have their own drawbacks. In this paper, we propose a novel architecture, LFI-CAM***(Learning Feature Importance Class Activation Mapping), which is trainable for image classification and visual explanation in an end-to-end manner. LFI-CAM generates attention map for visual explanation during forward propagation, and simultaneously uses attention map to improve classification performance through the attention mechanism. Feature Importance Network (FIN) focuses on learning the feature importance instead of directly learning the attention map to obtain a more reliable and consistent attention map. We confirmed that LFI-CAM is optimized not only by learning the feature importance but also by enhancing the backbone feature representation to focus more on important features of the input image. Experiments show that LFI-CAM outperforms baseline models’ accuracy on classification tasks as well as significantly improves on previous works in terms of attention map quality and stability over different hyper-parameters.

A B-hypergraph consisting of nodes and directed hyperedges is a generalization of the directed graph. A directed hyperedge in the B-hypergraph represents a relation from a set of source nodes to a single destination node. We suggest one possible definition of betweenness centrality (BC) in B-hypergraphs, called Participation-based BC (PBC). A PBC score of a node is computed based on the number of the shortest paths in which the node participates. This score can be expressed in terms of dependency on the set of its outgoing hyperedges. In this article, we focus on developing efficient computation algorithms for PBC. We first present an algorithm called ePBC for computing exact PBC scores of nodes, which has a cubic-time complexity. This algorithm, however, can be used for only small-sized B-hypergraphs because of its cubic-time complexity, so we propose linearized PBC ( ℓ PBC) that is an approximation method of ePBC. ℓ PBC that comes with a guaranteed upper bound on its error, uses a linearization of dependency on a set of hyperedges. ℓ PBC improves the computing time of ePBC by an order of magnitude (i.e., it requires a quadratic time) while maintaining a high accuracy. ℓ PBC works well on small to medium-sized B-hypergraphs, but is not scalable enough for a very large B-hypergraph with more than a million hyperedges. To cope with such a very large B-hypergraph, we propose a very fast heuristic sampling-based method called sampling-based ℓ PBC (s ℓ PBC). We show through extensive experiments that ℓ PBC and s ℓ PBC can efficiently estimate PBC scores in various real-world B-hypergraphs with a reasonably good precision@ k . The experimental results show that s ℓ PBC works efficiently even for a very large B-hypergraph.

Arbitrary style transfer aims to synthesize a content image with the style of an image to create a third image that has never been seen before. Recent arbitrary style transfer algorithms find it challenging to balance the content structure and the style patterns. Moreover, simultaneously maintaining the global and local style patterns is difficult due to the patch-based mechanism. In this paper, we introduce a novel style-attentional network (SANet) that efficiently and flexibly integrates the local style patterns according to the semantic spatial distribution of the content image. A new identity loss function and multi-level feature embeddings enable our SANet and decoder to preserve the content structure as much as possible while enriching the style patterns. Experimental results demonstrate that our algorithm synthesizes stylized images in real-time that are higher in quality than those produced by the state-of-the-art algorithms.

The directed hypergraph (especially B-hypergraph) has hyperedges that represent relations of a set of source nodes to a single target node. Author-cited networks and cellular signaling pathways can be modeled as a B-hypergraph. In this paper every source node of a hyperedge in the shortest path p in a B-hypergraph is considered a participant of p. We propose a betweenness centrality in the B-hypergraph that measures the number of shortest paths in which a node participates. The algorithm for computing the approximated betweenness centrality scores is also proposed. Through various performance experiments such as attack robustness and reachability tests, we show that our proposed betweenness centrality is a more appropriate measure in real-world B-hypergraph applications than ordinary betweenness centrality.

We present an efficient deterministic hypothesis generation algorithm for robust fitting of multiple structures based on the maximum feasible subsystem (MaxFS) framework. Despite its advantage, a global optimization method such as MaxFS has two main limitations for geometric model fitting. First, its performance is much influenced by the user-specified inlier scale. Second, it is computationally inefficient for large data. The presented algorithm, called iterative MaxFS with inlier scale (IMaxFS-ISE), iteratively estimates model parameters and inlier scale and also overcomes the second limitation by reducing data for the MaxFS problem. The IMaxFS-ISE algorithm generates hypotheses only with top-n ranked subsets based on matching scores and data fitting residuals. This reduction of data for the MaxFS problem makes the algorithm computationally realistic. A sequential "fitting-and-removing" procedure is repeated until overall energy function does not decrease. Experimental results demonstrate that our method can generate more reliable and consistent hypotheses than random sampling-based methods for estimating multiple structures from data with many outliers.