Recent works suggest that transformer models are capable of multi-tasking on diverse NLP tasks and adapting to new tasks efficiently. However, the potential of these multi-task models may be limited as they use the same set of parameters for all tasks. In contrast, humans tackle tasks in a more flexible way, by making proper presumptions on what skills and knowledge are relevant and executing only the necessary computations. Inspired by this, we propose to use task-level mixture-of-expert models, which has a collection of transformer layers (i.e., experts) and a router component that chooses from these experts dynamically and flexibly. We find that these models help improve the average performance gain (ARG) metric by 2.6% when adapting to unseen tasks in the few-shot setting and by 5.6% in the zero-shot generalization setting. Further, we show that the learned routing decisions partly rediscover human categorization of NLP tasks -- certain experts are strongly associated with extractive tasks, some with classification tasks, and some with tasks requiring world knowledge.

In natural language processing, extreme multi-label text classification is an emerging but essential task. The problem of extreme multi-label text classification (XMTC) is to recall some of the most relevant labels for a text from an extremely large label set. Large-scale pre-trained models have brought a new trend to this problem. Though the large-scale pre-trained models have made significant achievements on this problem, the valuable fine-tuned methods have yet to be studied. Though label semantics have been introduced in XMTC, the vast semantic gap between texts and labels has yet to gain enough attention. This paper builds a new guide network (GUDN) to help fine-tune the pre-trained model to instruct classification later. Furthermore, GUDN uses raw label semantics combined with a helpful label reinforcement strategy to effectively explore the latent space between texts and labels, narrowing the semantic gap, which can further improve predicted accuracy. Experimental results demonstrate that GUDN outperforms state-of-the-art methods on Eurlex-4k and has competitive results on other popular datasets. In an additional experiment, we investigated the input lengths' influence on the Transformer-based model's accuracy. Our source code is released at https://t.hk.uy/aFSH.

The paper analyses and serves with a positioning of various error measures applied in neural network training and identifies that there is no best of measure, although there is a set of measures with changing superiorities in different learning situations. An outstanding, remarkable measure called $E_{Exp}$ published by Silva and his research partners represents a research direction to combine more measures successfully with fixed importance weighting during learning. The main idea of the paper is to go far beyond and to integrate this relative importance into the neural network training algorithm(s) realized through a novel error measure called $E_{ExpAbs}$. This approach is included into the Levenberg-Marquardt training algorithm, so, a novel version of it is also introduced, resulting a self-adaptive, dynamic learning algorithm. This dynamism does not has positive effects on the resulted model accuracy only, but also on the training process itself. The described comprehensive algorithm tests proved that the proposed, novel algorithm integrates dynamically the two big worlds of statistics and information theory that is the key novelty of the paper.

Under the Autonomous Mobile Clinics (AMCs) initiative, we are developing, open sourcing, and standardizing health AI technologies to enable healthcare access in least developed countries (LDCs). We deem AMCs as the next generation of health care delivery platforms, whereas health AI engines are applications on these platforms, similar to how various applications expand the usage scenarios of smart phones. Facing the recent global monkeypox outbreak, in this article, we introduce AICOM-MP, an AI-based monkeypox detector specially aiming for handling images taken from resource-constrained devices. Compared to existing AI-based monkeypox detectors, AICOM-MP has achieved state-of-the-art (SOTA) performance. We have hosted AICOM-MP as a web service to allow universal access to monkeypox screening technology. We have also open sourced both the source code and the dataset of AICOM-MP to allow health AI professionals to integrate AICOM-MP into their services. Also, through the AICOM-MP project, we have generalized a methodology of developing health AI technologies for AMCs to allow universal access even in resource-constrained environments.

Artificial intelligence is becoming increasingly significant in facility management. In particular, predictive maintenance is a significant application of AI in this domain. Predictive maintenance is the process of using data to predict when equipment will fail and needs to be repaired or replaced. This is significant because it can help prevent equipment failures, which can lead to disruptions in service. With the pressure on organizations to do more with less, Facilities Management (FM) must challenge itself to be a strategic business enabler.

The attendees learned what JAX is and its fundamental yet unique features, which make it efficient to use when executing deep learning workloads. After that, they started training their first JAX-powered deep learning model. TFUG Taipei hosted Python JAX Image classification and helped people learn JAX and how to use it in Colab. They shared knowledge about the difference between JAX and Numpy, the advantages of JAX, and how to use it in Colab. Introduction to JAX by ML GDE João Araújo (Brazil) shared the basics of JAX in Deep Learning Indaba 2022.

How to improve the accuracy of emotion recognition has become a primary issue. In recent years, with the continuous development of artificial intelligence technology, human-computer interaction has become the focus of research in the field of information science. As one of the critical technologies to realize human-computer interaction, emotion recognition has gradually received a lot of attention from researchers. At present, most of the research works on emotion recognition are based on single-modal, such as facial expressions [1-3], body movements [4-5] and speech text [6-7]. However, emotion recognition based on unimodal often has limitations and, in most cases, could only reflect a portion of human emotional expression. Multimodal emotion recognition can link individual unimodal channels and use the feature complementarity between channels to combine multiple information to determine the emotional state. Studies have shown that the multimodal emotion recognition approach has better performance than unimodal emotion judgment in most cases [8]. The difficulty of multimodal recognition is not only to control the internal information of individual modality (Intra-modality), but also to complement the interactive features between individual modalities (Inter-modality). It has been extensively studied by scholars, such as Tensor Fusion Network (TFN) proposed by Zadeh et al [9], Polynomial Tensor Pooling (PTP) proposed by Hou et al [10], and Memory Fusion Network (MFN) presented by Zadeh et al [11].

Zero-shot entity retrieval, aiming to link mentions to candidate entities under the zero-shot setting, is vital for many tasks in Natural Language Processing. Most existing methods represent mentions/entities via the sentence embeddings of corresponding context from the Pre-trained Language Model. However, we argue that such coarse-grained sentence embeddings can not fully model the mentions/entities, especially when the attention scores towards mentions/entities are relatively low. In this work, we propose GER, a \textbf{G}raph enhanced \textbf{E}ntity \textbf{R}etrieval framework, to capture more fine-grained information as complementary to sentence embeddings. We extract the knowledge units from the corresponding context and then construct a mention/entity centralized graph. Hence, we can learn the fine-grained information about mention/entity by aggregating information from these knowledge units. To avoid the graph information bottleneck for the central mention/entity node, we construct a hierarchical graph and design a novel Hierarchical Graph Attention Network~(HGAN). Experimental results on popular benchmarks demonstrate that our proposed GER framework performs better than previous state-of-the-art models. The code has been available at https://github.com/wutaiqiang/GER-WSDM2023.

Reliable and accurate registration of patient-specific brain magnetic resonance imaging (MRI) scans containing pathologies is challenging due to tissue appearance changes. This paper describes our contribution to the Registration of the longitudinal brain MRI task of the Brain Tumor Sequence Registration Challenge 2022 (BraTS-Reg 2022). We developed an enhanced unsupervised learning-based method that extends the iRegNet. In particular, incorporating an unsupervised learning-based paradigm as well as several minor modifications to the network pipeline, allows the enhanced iRegNet method to achieve respectable results. Experimental findings show that the enhanced self-supervised model is able to improve the initial mean median registration absolute error (MAE) from 8.20 (7.62) mm to the lowest value of 3.51 (3.50) for the training set while achieving an MAE of 2.93 (1.63) mm for the validation set. Additional qualitative validation of this study was conducted through overlaying pre-post MRI pairs before and after the de-formable registration. The proposed method scored 5th place during the testing phase of the MICCAI BraTS-Reg 2022 challenge. The docker image to reproduce our BraTS-Reg submission results will be publicly available.

The increasing availability of geometric data has motivated the need for information processing over non-Euclidean domains modeled as manifolds. The building block for information processing architectures with desirable theoretical properties such as invariance and stability is convolutional filtering. Manifold convolutional filters are defined from the manifold diffusion sequence, constructed by successive applications of the Laplace-Beltrami operator to manifold signals. However, the continuous manifold model can only be accessed by sampling discrete points and building an approximate graph model from the sampled manifold. Effective linear information processing on the manifold requires quantifying the error incurred when approximating manifold convolutions with graph convolutions. In this paper, we derive a non-asymptotic error bound for this approximation, showing that convolutional filtering on the sampled manifold converges to continuous manifold filtering. Our findings are further demonstrated empirically on a problem of navigation control.