Researchers have long tried to minimize training costs in deep learning while maintaining strong generalization across diverse datasets. Emerging research on dataset distillation aims to reduce training costs by creating a small synthetic set that contains the information of a larger real dataset and ultimately achieves test accuracy equivalent to a model trained on the whole dataset. Unfortunately, the synthetic data generated by previous methods are not guaranteed to distribute and discriminate as well as the original training data, and they incur significant computational costs. Despite promising results, there still exists a significant performance gap between models trained on condensed synthetic sets and those trained on the whole dataset. In this paper, we address these challenges using efficient Dataset Distillation with Attention Matching (DataDAM), achieving state-of-the-art performance while reducing training costs. Specifically, we learn synthetic images by matching the spatial attention maps of real and synthetic data generated by different layers within a family of randomly initialized neural networks. Our method outperforms the prior methods on several datasets, including CIFAR10/100, TinyImageNet, ImageNet-1K, and subsets of ImageNet-1K across most of the settings, and achieves improvements of up to 6.5% and 4.1% on CIFAR100 and ImageNet-1K, respectively. We also show that our high-quality distilled images have practical benefits for downstream applications, such as continual learning and neural architecture search.

Twitter user geolocation detection can inform and benefit a range of downstream geospatial tasks such as event and venue recommendation, local search, and crisis planning and response. In this paper, we take into account user shared tweets as well as their social network, and run extensive comparative studies to systematically analyze the impact of a variety of language-based, network-based, and hybrid methods in predicting user geolocation. In particular, we evaluate different text representation methods to construct text views that capture the linguistic signals available in tweets that are specific to and indicative of geographical locations. In addition, we investigate a range of network-based methods, such as embedding approaches and graph neural networks, in predicting user geolocation based on user interaction network. Our findings provide valuable insights into the design of effective and efficient geolocation identification engines. Finally, our best model, called TF-MF, substantially outperforms state-of-the-art approaches under minimal supervision.