Passenger clustering based on travel records is essential for transportation operators. However, existing methods cannot easily cluster the passengers due to the hierarchical structure of the passenger trip information, namely: each passenger has multiple trips, and each trip contains multi-dimensional multi-mode information. Furthermore, existing approaches rely on an accurate specification of the clustering number to start, which is difficult when millions of commuters are using the transport systems on a daily basis. In this paper, we propose a novel Tensor Dirichlet Process Multinomial Mixture model (Tensor-DPMM), which is designed to preserve the multi-mode and hierarchical structure of the multi-dimensional trip information via tensor, and cluster them in a unified one-step manner. The model also has the ability to determine the number of clusters automatically by using the Dirichlet Process to decide the probabilities for a passenger to be either assigned in an existing cluster or to create a new cluster: This allows our model to grow the clusters as needed in a dynamic manner. Finally, existing methods do not consider spatial semantic graphs such as geographical proximity and functional similarity between the locations, which may cause inaccurate clustering. To this end, we further propose a variant of our model, namely the Tensor-DPMM with Graph. For the algorithm, we propose a tensor Collapsed Gibbs Sampling method, with an innovative step of "disband and relocating", which disbands clusters with too small amount of members and relocates them to the remaining clustering. This avoids uncontrollable growing amounts of clusters. A case study based on Hong Kong metro passenger data is conducted to demonstrate the automatic process of learning the number of clusters, and the learned clusters are better in within-cluster compactness and cross-cluster separateness.

In recent years, the development of autonomous electric vehicles (AEVs) has gained significant attention from researchers and engineers worldwide. AEVs are expected to revolutionize the way we commute and transport goods, offering safer and more efficient solutions to our transportation needs.

Despite the impressive performance recently achieved by automatic speech recognition (ASR), we observe two primary challenges that hinder its broader applications: (1) The difficulty of introducing scalability into the model to support more languages with limited training, inference, and storage overhead; (2) The low-resource adaptation ability that enables effective low-resource adaptation while avoiding over-fitting and catastrophic forgetting issues. Inspired by recent findings, we hypothesize that we can address the above challenges with modules widely shared across languages. To this end, we propose an ASR framework, dubbed \METHODNS, that, \textit{for the first time}, simultaneously achieves strong multilingual scalability and low-resource adaptation ability thanks to its modularize-then-assemble strategy. Specifically, \METHOD learns a small set of generalizable sub-modules and adaptively assembles them for different languages to reduce the multilingual overhead and enable effective knowledge transfer for low-resource adaptation. Extensive experiments and visualizations demonstrate that \METHOD can effectively discover language similarity and improve multilingual and low-resource ASR performance over state-of-the-art (SOTA) methods, e.g., under multilingual-ASR, our framework achieves a 0.13$\sim$2.41 lower character error rate (CER) with 30\% smaller inference overhead over SOTA solutions on multilingual ASR and a comparable CER, with nearly 50 times fewer trainable parameters over SOTA solutions on low-resource tuning, respectively.

The capabilities and adoption of deep neural networks (DNNs) grow at an exhilarating pace: Vision models accurately classify human actions in videos and identify cancerous tissue in medical scans as precisely than human experts; large language models answer wide-ranging questions, generate code, and write prose, becoming the topic of everyday dinner-table conversations. Even though their uses are exhilarating, the continually increasing model sizes and computational complexities have a dark side. The economic cost and negative environmental externalities of training and serving models is in evident disharmony with financial viability and climate action goals. Instead of pursuing yet another increase in predictive performance, this dissertation is dedicated to the improvement of neural network efficiency. Specifically, a core contribution addresses the efficiency aspects during online inference. Here, the concept of Continual Inference Networks (CINs) is proposed and explored across four publications. CINs extend prior state-of-the-art methods developed for offline processing of spatio-temporal data and reuse their pre-trained weights, improving their online processing efficiency by an order of magnitude. These advances are attained through a bottom-up computational reorganization and judicious architectural modifications. The benefit to online inference is demonstrated by reformulating several widely used network architectures into CINs, including 3D CNNs, ST-GCNs, and Transformer Encoders. An orthogonal contribution tackles the concurrent adaptation and computational acceleration of a large source model into multiple lightweight derived models. Drawing on fusible adapter networks and structured pruning, Structured Pruning Adapters achieve superior predictive accuracy under aggressive pruning using significantly fewer learned weights compared to fine-tuning with pruning.

Retrieval-augmented language models (LMs) have received much attention recently. However, typically the retriever is not trained jointly as a native component of the LM, but added to an already-pretrained LM, which limits the ability of the LM and the retriever to adapt to one another. In this work, we propose the Retrieval-Pretrained Transformer (RPT), an architecture and training procedure for jointly training a retrieval-augmented LM from scratch for the task of modeling long texts. Given a recently generated text chunk in a long document, the LM computes query representations, which are then used to retrieve earlier chunks in the document, located potentially tens of thousands of tokens before. Information from retrieved chunks is fused into the LM representations to predict the next target chunk. We train the retriever component with a semantic objective, where the goal is to retrieve chunks that increase the probability of the next chunk, according to a reference LM. We evaluate RPT on four long-range language modeling tasks, spanning books, code, and mathematical writing, and demonstrate that RPT improves retrieval quality and subsequently perplexity across the board compared to strong baselines.

The advent of AI driven large language models (LLMs) have stirred discussions about their role in qualitative research. Some view these as tools to enrich human understanding, while others perceive them as threats to the core values of the discipline. This study aimed to compare and contrast the comprehension capabilities of humans and LLMs. We conducted an experiment with small sample of Alexa app reviews, initially classified by a human analyst. LLMs were then asked to classify these reviews and provide the reasoning behind each classification. We compared the results with human classification and reasoning. The research indicated a significant alignment between human and ChatGPT 3.5 classifications in one third of cases, and a slightly lower alignment with GPT4 in over a quarter of cases. The two AI models showed a higher alignment, observed in more than half of the instances. However, a consensus across all three methods was seen only in about one fifth of the classifications. In the comparison of human and LLMs reasoning, it appears that human analysts lean heavily on their individual experiences. As expected, LLMs, on the other hand, base their reasoning on the specific word choices found in app reviews and the functional components of the app itself. Our results highlight the potential for effective human LLM collaboration, suggesting a synergistic rather than competitive relationship. Researchers must continuously evaluate LLMs role in their work, thereby fostering a future where AI and humans jointly enrich qualitative research.

This work introduces an attention mechanism for image classifiers and the corresponding deep neural network (DNN) architecture, dubbed ISNet. During training, the ISNet uses segmentation targets to learn how to find the image's region of interest and concentrate its attention on it. The proposal is based on a novel concept, background relevance minimization in LRP explanation heatmaps. It can be applied to virtually any classification neural network architecture, without any extra computational cost at run-time. Capable of ignoring the background, the resulting single DNN can substitute the common pipeline of a segmenter followed by a classifier, being faster and lighter. After injecting synthetic bias in images' backgrounds (in diverse applications), we compare the ISNet to multiple state-of-the-art neural networks, and quantitatively demonstrate its superior capacity of minimizing the bias influence over the classifier decisions. The tasks of COVID-19 and tuberculosis detection in chest X-rays commonly employ mixed training databases, which naturally foster background bias and shortcut learning. By focusing on lungs, the ISNet reduced shortcut learning, leading to significantly superior generalization to external (out-of-distribution) test datasets. ISNet presents an accurate, fast, and light methodology to ignore backgrounds and improve generalization.

The ongoing Russo-Ukrainian conflict has been a subject of intense media coverage worldwide. Understanding the global narrative surrounding this topic is crucial for researchers that aim to gain insights into its multifaceted dimensions. In this paper, we present a novel dataset that focuses on this topic by collecting and processing tweets posted by news or media companies on social media across the globe. We collected tweets from February 2022 to May 2023 to acquire approximately 1.5 million tweets in 60 different languages. Each tweet in the dataset is accompanied by processed tags, allowing for the identification of entities, stances, concepts, and sentiments expressed. The availability of the dataset serves as a valuable resource for researchers aiming to investigate the global narrative surrounding the ongoing conflict from various aspects such as who are the prominent entities involved, what stances are taken, where do these stances originate, and how are the different concepts related to the event portrayed.

There are many complex and rich morphological subtleties in the Arabic language, which are very useful when analyzing traditional Arabic texts, especially in the historical and religious contexts, and help in understanding the meaning of the texts. Vocabulary separation means separating the word into different parts such as root and affix. In the morphological datasets, the variety of labels and the number of data samples helps to evaluate the morphological methods. In this paper, we present a benchmark data set for evaluating the methods of separating Arabic words which include about 223,690 words from the book of Sharia alIslam, which have been labeled by experts. In terms of the volume and variety of words, this dataset is superior to other existing data sets, and as far as we know, there are no Arabic Hadith Domain texts. To evaluate the dataset, we applied different methods such as Farasa, Camel, Madamira, and ALP to the dataset and we reported the annotation quality through four evaluation methods.

Simultaneous localisation and mapping (SLAM) play a vital role in autonomous robotics. Robotic platforms are often resource-constrained, and this limitation motivates resource-efficient SLAM implementations. While sparse visual SLAM algorithms offer good accuracy for modest hardware requirements, even these more scalable sparse approaches face limitations when applied to large-scale and long-term scenarios. A contributing factor is that the point clouds resulting from SLAM are inefficient to use and contain significant redundancy. This paper proposes the use of subset selection algorithms to reduce the map produced by sparse visual SLAM algorithms. Information-theoretic techniques have been applied to simpler related problems before, but they do not scale if applied to the full visual SLAM problem. This paper proposes a number of novel information\hyp{}theoretic utility functions for map point selection and optimises these functions using greedy algorithms. The reduced maps are evaluated using practical data alongside an existing visual SLAM implementation (ORB-SLAM 2). Approximate selection techniques proposed in this paper achieve trajectory accuracy comparable to an offline baseline while being suitable for online use. These techniques enable the practical reduction of maps for visual SLAM with competitive trajectory accuracy. Results also demonstrate that SLAM front-end performance can significantly impact the performance of map point selection. This shows the importance of testing map point selection with a front-end implementation. To exploit this, this paper proposes an approach that includes a model of the front-end in the utility function when additional information is available. This approach outperforms alternatives on applicable datasets and highlights future research directions.