Natural language processing (NLP) sees rich mobile applications. To support various language understanding tasks, a foundation NLP model is often fine-tuned in a federated, privacy-preserving setting (FL). This process currently relies on at least hundreds of thousands of labeled training samples from mobile clients; yet mobile users often lack willingness or knowledge to label their data. Such an inadequacy of data labels is known as a few-shot scenario; it becomes the key blocker for mobile NLP applications. For the first time, this work investigates federated NLP in the few-shot scenario (FedFSL). By retrofitting algorithmic advances of pseudo labeling and prompt learning, we first establish a training pipeline that delivers competitive accuracy when only 0.05% (fewer than 100) of the training data is labeled and the remaining is unlabeled. To instantiate the workflow, we further present a system FFNLP, addressing the high execution cost with novel designs. (1) Curriculum pacing, which injects pseudo labels to the training workflow at a rate commensurate to the learning progress; (2) Representational diversity, a mechanism for selecting the most learnable data, only for which pseudo labels will be generated; (3) Co-planning of a model's training depth and layer capacity. Together, these designs reduce the training delay, client energy, and network traffic by up to 46.0$\times$, 41.2$\times$ and 3000.0$\times$, respectively. Through algorithm/system co-design, FFNLP demonstrates that FL can apply to challenging settings where most training samples are unlabeled.
Detecting unsafe driving states, such as stress, drowsiness, and fatigue, is an important component of ensuring driving safety and an essential prerequisite for automatic intervention systems in vehicles. These concerning conditions are primarily connected to the driver's low or high arousal levels. In this study, we describe a framework for processing multimodal physiological time-series from wearable sensors during driving and locating points of prominent change in drivers' physiological arousal state. These points of change could potentially indicate events that require just-in-time intervention. We apply time-series segmentation on heart rate and breathing rate measurements and quantify their robustness in capturing change points in electrodermal activity, treated as a reference index for arousal, as well as on self-reported stress ratings, using three public datasets. Our experiments demonstrate that physiological measures are veritable indicators of change points of arousal and perform robustly across an extensive ablation study.
Security issues are threatened in various types of networks, especially in the Internet of Things (IoT) environment that requires early detection. IoT is the network of real-time devices like home automation systems and can be controlled by open-source android devices, which can be an open ground for attackers. Attackers can access the network, initiate a different kind of security breach, and compromises network control. Therefore, timely detecting the increasing number of sophisticated malware attacks is the challenge to ensure the credibility of network protection. In this regard, we have developed a new malware detection framework, Deep Squeezed-Boosted and Ensemble Learning (DSBEL), comprised of novel Squeezed-Boosted Boundary-Region Split-Transform-Merge (SB-BR-STM) CNN and ensemble learning. The proposed S.T.M. block employs multi-path dilated convolutional, Boundary, and regional operations to capture the homogenous and heterogeneous global malicious patterns. Moreover, diverse feature maps are achieved using transfer learning and multi-path-based squeezing and boosting at initial and final levels to learn minute pattern variations. Finally, the boosted discriminative features are extracted from the developed deep SB-BR-STM CNN and provided to the ensemble classifiers (SVM, M.L.P., and AdaboostM1) to improve the hybrid learning generalization. The performance analysis of the proposed DSBEL framework and SB-BR-STM CNN against the existing techniques have been evaluated by the IOT_Malware dataset on standard performance measures. Evaluation results show progressive performance as 98.50% accuracy, 97.12% F1-Score, 91.91% MCC, 95.97 % Recall, and 98.42 % Precision. The proposed malware analysis framework is helpful for the timely detection of malicious activity and suggests future strategies.
Although pre-trained language models (PLMs) have shown impressive performance by text-only self-supervised training, they are found lack of visual semantics or commonsense, e.g., sizes, shapes, and colors of commonplace objects. Existing solutions often rely on explicit images for visual knowledge augmentation (requiring time-consuming retrieval or generation), and they also conduct the augmentation for the whole input text, without considering whether it is actually needed in specific inputs or tasks. To address these issues, we propose a novel visually-augmented fine-tuning approach that can be generally applied to various PLMs or NLP tasks, without using any retrieved or generated images, namely VAWI. Specifically, we first identify the visually-hungry words (VH-words) from input text via a token selector, where three different methods have been proposed, including syntax-, attention- and learning-based strategies. Then, we adopt a fixed CLIP text encoder to generate the visually-augmented representations of these VH-words. As it has been pre-trained by vision-language alignment task on the large-scale corpus, it is capable of injecting visual semantics into the aligned text representations. Finally, the visually-augmented features will be fused and transformed into the pre-designed visual prompts based on VH-words, which can be inserted into PLMs to enrich the visual semantics in word representations. We conduct extensive experiments on ten NLP tasks, i.e., GLUE benchmark, CommonsenseQA, CommonGen, and SNLI-VE. Experimental results show that our approach can consistently improve the performance of BERT, RoBERTa, BART, and T5 at different scales, and outperform several competitive baselines significantly. Our codes and data are publicly available at~\url{https://github.com/RUCAIBox/VAWI}.
Reinforcement learning (RL) gained considerable attention by creating decision-making agents that maximize rewards received from fully observable environments. However, many real-world problems are partially or noisily observable by nature, where agents do not receive the true and complete state of the environment. Such problems are formulated as partially observable Markov decision processes (POMDPs). Some studies applied RL to POMDPs by recalling previous decisions and observations or inferring the true state of the environment from received observations. Nevertheless, aggregating observations and decisions over time is impractical for environments with high-dimensional continuous state and action spaces. Moreover, so-called inference-based RL approaches require large number of samples to perform well since agents eschew uncertainty in the inferred state for the decision-making. Active inference is a framework that is naturally formulated in POMDPs and directs agents to select decisions by minimising expected free energy (EFE). This supplies reward-maximising (exploitative) behaviour in RL, with an information-seeking (exploratory) behaviour. Despite this exploratory behaviour of active inference, its usage is limited to discrete state and action spaces due to the computational difficulty of the EFE. We propose a unified principle for joint information-seeking and reward maximization that clarifies a theoretical connection between active inference and RL, unifies active inference and RL, and overcomes their aforementioned limitations. Our findings are supported by strong theoretical analysis. The proposed framework's superior exploration property is also validated by experimental results on partial observable tasks with high-dimensional continuous state and action spaces. Moreover, the results show that our model solves reward-free problems, making task reward design optional.
Large pre-trained language models (LPLM) have shown spectacular success when fine-tuned on downstream supervised tasks. Yet, it is known that their performance can drastically drop when there is a distribution shift between the data used during training and that used at inference time. In this paper we focus on data distributions that naturally change over time and introduce four new REDDIT datasets, namely the WALLSTREETBETS, ASKSCIENCE, THE DONALD, and POLITICS sub-reddits. First, we empirically demonstrate that LPLM can display average performance drops of about 88% (in the best case!) when predicting the popularity of future posts from sub-reddits whose topic distribution changes with time. We then introduce a simple methodology that leverages neural variational dynamic topic models and attention mechanisms to infer temporal language model representations for regression tasks. Our models display performance drops of only about 40% in the worst cases (2% in the best ones) when predicting the popularity of future posts, while using only about 7% of the total number of parameters of LPLM and providing interpretable representations that offer insight into real-world events, like the GameStop short squeeze of 2021
We introduce DiffRF, a novel approach for 3D radiance field synthesis based on denoising diffusion probabilistic models. While existing diffusion-based methods operate on images, latent codes, or point cloud data, we are the first to directly generate volumetric radiance fields. To this end, we propose a 3D denoising model which directly operates on an explicit voxel grid representation. However, as radiance fields generated from a set of posed images can be ambiguous and contain artifacts, obtaining ground truth radiance field samples is non-trivial. We address this challenge by pairing the denoising formulation with a rendering loss, enabling our model to learn a deviated prior that favours good image quality instead of trying to replicate fitting errors like floating artifacts. In contrast to 2D-diffusion models, our model learns multi-view consistent priors, enabling free-view synthesis and accurate shape generation. Compared to 3D GANs, our diffusion-based approach naturally enables conditional generation such as masked completion or single-view 3D synthesis at inference time.
Deep learning models require an enormous amount of data for training. However, recently there is a shift in machine learning from model-centric to data-centric approaches. In data-centric approaches, the focus is to refine and improve the quality of the data to improve the learning performance of the models rather than redesigning model architectures. In this paper, we propose CLIP i.e., Curriculum Learning with Iterative data Pruning. CLIP combines two data-centric approaches i.e., curriculum learning and dataset pruning to improve the model learning accuracy and convergence speed. The proposed scheme applies loss-aware dataset pruning to iteratively remove the least significant samples and progressively reduces the size of the effective dataset in the curriculum learning training. Extensive experiments performed on crowd density estimation models validate the notion behind combining the two approaches by reducing the convergence time and improving generalization. To our knowledge, the idea of data pruning as an embedded process in curriculum learning is novel.
Understanding pedestrian behavior patterns is a key component to building autonomous agents that can navigate among humans. We seek a learned dictionary of pedestrian behavior to obtain a semantic description of pedestrian trajectories. Supervised methods for dictionary learning are impractical since pedestrian behaviors may be unknown a priori and the process of manually generating behavior labels is prohibitively time consuming. We instead utilize a novel, unsupervised framework to create a taxonomy of pedestrian behavior observed in a specific space. First, we learn a trajectory latent space that enables unsupervised clustering to create an interpretable pedestrian behavior dictionary. We show the utility of this dictionary for building pedestrian behavior maps to visualize space usage patterns and for computing the distributions of behaviors. We demonstrate a simple but effective trajectory prediction by conditioning on these behavior labels. While many trajectory analysis methods rely on RNNs or transformers, we develop a lightweight, low-parameter approach and show results comparable to SOTA on the ETH and UCY datasets.
We developed an automated method for sunspot detection using digital white-light solar images to achieve a performance similar to that of visual drawing observations in sunspot counting. To identify down to small, isolated spots correctly, we pay special attention to the accurate derivation of the quiet-disk component of the Sun, which is used as a reference to identify sunspots using a threshold. This threshold is determined using an adaptive method to process images obtained under various conditions. To eliminate the seeing effect, our method can process multiple images taken within a short time. We applied the developed method to digital images captured at three sites and compared the detection results with those of visual observations. We conclude that the proposed sunspot detection method has a similar performance to that of visual observation. This method can be widely used by public observatories and amateurs as well as professional observatories as an alternative to hand-drawn visual observation for sunspot counting.