Spatio-temporal forecasting is crucial in various fields and requires a careful balance between identifying subtle patterns and filtering out noise. Vector quantization (VQ) appears well-suited for this purpose, as it quantizes input vectors into a set of codebook vectors or patterns. Although VQ has shown promise in various computer vision tasks, it surprisingly falls short in enhancing the accuracy of spatio-temporal forecasting. We attribute this to two main issues: inaccurate optimization due to non-differentiability and limited representation power caused by hard-VQ structure. To tackle these challenges, we introduce Differentiable Sparse Soft-Vector Quantization (SVQ), the first VQ method to enhance spatio-temporal forecasting. SVQ balances detail preservation with noise reduction, offering full differentiability and a solid foundation in sparse regression. The method employs a two-layer MLP and an extensive codebook to streamline the sparse regression process, significantly cutting computational costs while simplifying training and improving performance. Empirical studies on five spatio-temporal benchmark datasets show SVQ achieves state-of-the-art results, including a 7.9% improvement on the WeatherBench-S temperature dataset and an average mean absolute error reduction of 9.4% in video prediction benchmarks (Human3.6M, KTH, and KittiCaltech), along with a 17.3% enhancement in image quality (measured by LPIPS).

Weather forecasting has long posed a significant challenge for humanity. While recent AI-based models have surpassed traditional numerical weather prediction (NWP) methods in global forecasting tasks, overfitting remains a critical issue due to the limited availability of real-world weather data spanning only a few decades. Unlike fields like computer vision or natural language processing, where data abundance can mitigate overfitting, weather forecasting demands innovative strategies to address this challenge with existing data. In this paper, we explore pre-training methods for weather forecasting, finding that selecting an appropriately challenging pre-training task introduces locality bias, effectively mitigating overfitting and enhancing performance. We introduce Baguan, a novel data-driven model for medium-range weather forecasting, built on a Siamese Autoencoder pre-trained in a self-supervised manner and fine-tuned for different lead times. Experimental results show that Baguan outperforms traditional methods, delivering more accurate forecasts. Additionally, the pre-trained Baguan demonstrates robust overfitting control and excels in downstream tasks, such as subseasonal-to-seasonal (S2S) modeling and regional forecasting, after fine-tuning.

Accurate solar power forecasting is crucial to integrate photovoltaic plants into the electric grid, schedule and secure the power grid safety. This problem becomes more demanding for those newly installed solar plants which lack sufficient operational data. Current research predominantly relies on historical solar power data or numerical weather prediction in a single-modality format, ignoring the complementary information provided in different modalities. In this paper, we propose a multi-modality fusion framework to integrate historical power data, numerical weather prediction, and satellite images, significantly improving forecast performance. We introduce a vector quantized framework that aligns modalities with varying information densities, striking a balance between integrating sufficient information and averting model overfitting. Our framework demonstrates strong zero-shot forecasting capability, which is especially useful for those newly installed plants. Moreover, we collect and release a multi-modal solar power (MMSP) dataset from real-world plants to further promote the research of multi-modal solar forecasting algorithms. Our extensive experiments show that our model not only operates with robustness but also boosts accuracy in both zero-shot forecasting and scenarios rich with training data, surpassing leading models. We have incorporated it into our eForecaster platform and deployed it for more than 300 solar plants with a total capacity of over 15GW. Our code and dataset are accessible at https://github.com/DAMO-DI-ML/FusionSF.git.

Transformer-based models have emerged as promising tools for time series forecasting. However, these models cannot make accurate prediction for long input time series. On the one hand, they failed to capture long-range dependency within time series data. On the other hand, the long input sequence usually leads to large model size and high time complexity. To address these limitations, we present GCformer, which combines a structured global convolutional branch for processing long input sequences with a local Transformer-based branch for capturing short, recent signals. A cohesive framework for a global convolution kernel has been introduced, utilizing three distinct parameterization methods. The selected structured convolutional kernel in the global branch has been specifically crafted with sublinear complexity, thereby allowing for the efficient and effective processing of lengthy and noisy input signals. Empirical studies on six benchmark datasets demonstrate that GCformer outperforms state-of-the-art methods, reducing MSE error in multivariate time series benchmarks by 4.38% and model parameters by 61.92%. In particular, the global convolutional branch can serve as a plug-in block to enhance the performance of other models, with an average improvement of 31.93%, including various recently published Transformer-based models. Our code is publicly available at https://github.com/Yanjun-Zhao/GCformer.

Time series anomaly detection is critical for a wide range of applications. It aims to identify deviant samples from the normal sample distribution in time series. The most fundamental challenge for this task is to learn a representation map that enables effective discrimination of anomalies. Reconstruction-based methods still dominate, but the representation learning with anomalies might hurt the performance with its large abnormal loss. On the other hand, contrastive learning aims to find a representation that can clearly distinguish any instance from the others, which can bring a more natural and promising representation for time series anomaly detection. In this paper, we propose DCdetector, a multi-scale dual attention contrastive representation learning model. DCdetector utilizes a novel dual attention asymmetric design to create the permutated environment and pure contrastive loss to guide the learning process, thus learning a permutation invariant representation with superior discrimination abilities. Extensive experiments show that DCdetector achieves state-of-the-art results on multiple time series anomaly detection benchmark datasets. Code is publicly available at https://github.com/DAMO-DI-ML/KDD2023-DCdetector.

Time series anomaly detection is a challenging problem due to the complex temporal dependencies and the limited label data. Although some algorithms including both traditional and deep models have been proposed, most of them mainly focus on time-domain modeling, and do not fully utilize the information in the frequency domain of the time series data. In this paper, we propose a Time-Frequency analysis based time series Anomaly Detection model, or TFAD for short, to exploit both time and frequency domains for performance improvement. Besides, we incorporate time series decomposition and data augmentation mechanisms in the designed time-frequency architecture to further boost the abilities of performance and interpretability. Empirical studies on widely used benchmark datasets show that our approach obtains state-of-the-art performance in univariate and multivariate time series anomaly detection tasks.

Time series forecasting is a critical and challenging problem in many real applications. Recently, Transformer-based models prevail in time series forecasting due to their advancement in long-range dependencies learning. Besides, some models introduce series decomposition to further unveil reliable yet plain temporal dependencies. Unfortunately, few models could handle complicated periodical patterns, such as multiple periods, variable periods, and phase shifts in real-world datasets. Meanwhile, the notorious quadratic complexity of dot-product attentions hampers long sequence modeling. To address these challenges, we design an innovative framework Quaternion Transformer (Quatformer), along with three major components: 1). learning-to-rotate attention (LRA) based on quaternions which introduces learnable period and phase information to depict intricate periodical patterns. 2). trend normalization to normalize the series representations in hidden layers of the model considering the slowly varying characteristic of trend. 3). decoupling LRA using global memory to achieve linear complexity without losing prediction accuracy. We evaluate our framework on multiple real-world time series datasets and observe an average 8.1% and up to 18.5% MSE improvement over the best state-of-the-art baseline.

Time series analysis is ubiquitous and important in various areas, such as Artificial Intelligence for IT Operations (AIOps) in cloud computing, AI-powered Business Intelligence (BI) in E-commerce, Artificial Intelligence of Things (AIoT), etc. In real-world scenarios, time series data often exhibit complex patterns with trend, seasonality, outlier, and noise. In addition, as more time series data are collected and stored, how to handle the huge amount of data efficiently is crucial in many applications. We note that these significant challenges exist in various tasks like forecasting, anomaly detection, and fault cause localization. Therefore, how to design effective and efficient time series models for different tasks, which are robust to address the aforementioned challenging patterns and noise in real-world scenarios, is of great theoretical and practical interests. In this tutorial, we provide a comprehensive and organized tutorial on the state-of-the-art algorithms of robust time series analysis, ranging from traditional statistical methods to the most recent deep learning based methods. We will not only introduce the principle of time series algorithms, but also provide insights into how to apply them effectively in practical real-world industrial applications. Specifically, we organize the tutorial in a bottom-up framework. We first present preliminaries from different disciplines including robust statistics, signal processing, optimization, and deep learning. Then, we identify and discuss those most-frequently processing blocks in robust time series analysis, including periodicity detection, trend filtering, seasonal-trend decomposition, and time series similarity. Lastly, we discuss recent advances in multiple time series tasks including forecasting, anomaly detection, fault cause localization, and autoscaling, as well as practical lessons of large-scale time series applications from an industrial perspective.