Advancing Time Series Classification with Multimodal Language Modeling

For the advancements of time series classification, scrutinizing previous studies, most existing methods adopt a common learning-to-classify paradigm - a time series classifier model tries to learn the relation between sequence inputs and target label encoded by one-hot distribution. Although effective, this paradigm conceals two inherent limitations: (1) encoding target categories with one-hot distribution fails to reflect the comparability and similarity between labels, and (2) it is very difficult to learn transferable model across domains, which greatly hinder the development of universal serving paradigm. In this work, we propose InstructTime, a novel attempt to reshape time series classification as a learning-to-generate paradigm. Relying on the powerful generative capacity of the pre-trained language model, the core idea is to formulate the classification of time series as a multimodal understanding task, in which both task-specific instructions and raw time series are treated as multimodal inputs while the label information is represented by texts. To accomplish this goal, three distinct designs are developed in the InstructTime. Firstly, a time series discretization module is designed to convert continuous time series into a sequence of hard tokens to solve the inconsistency issue across modal inputs. To solve the modality representation gap issue, for one thing, we introduce an alignment projected layer before feeding the transformed token of time series into language models. For another, we highlight the necessity of auto-regressive pre-training across domains, which can facilitate the transferability of the language model and boost the generalization performance. Extensive experiments are conducted over benchmark datasets, whose results uncover the superior performance of InstructTime and the potential for a universal foundation model in time series classification.

Generative Pretrained Hierarchical Transformer for Time Series Forecasting

Recent efforts have been dedicated to enhancing time series forecasting accuracy by introducing advanced network architectures and self-supervised pretraining strategies. Nevertheless, existing approaches still exhibit two critical drawbacks. Firstly, these methods often rely on a single dataset for training, limiting the model's generalizability due to the restricted scale of the training data. Secondly, the one-step generation schema is widely followed, which necessitates a customized forecasting head and overlooks the temporal dependencies in the output series, and also leads to increased training costs under different horizon length settings. To address these issues, we propose a novel generative pretrained hierarchical transformer architecture for forecasting, named GPHT. There are two aspects of key designs in GPHT. On the one hand, we advocate for constructing a mixed dataset for pretraining our model, comprising various datasets from diverse data scenarios. This approach significantly expands the scale of training data, allowing our model to uncover commonalities in time series data and facilitating improved transfer to specific datasets. On the other hand, GPHT employs an auto-regressive forecasting approach under the channel-independent assumption, effectively modeling temporal dependencies in the output series. Importantly, no customized forecasting head is required, enabling a single model to forecast at arbitrary horizon settings. We conduct sufficient experiments on eight datasets with mainstream self-supervised pretraining models and supervised models. The results demonstrated that GPHT surpasses the baseline models across various fine-tuning and zero/few-shot learning settings in the traditional long-term forecasting task, providing support for verifying the feasibility of pretrained time series large models.

Towards Automatic Sampling of User Behaviors for Sequential Recommender Systems

Sequential recommender systems (SRS) have gained widespread popularity in recommendation due to their ability to effectively capture dynamic user preferences. One default setting in the current SRS is to uniformly consider each historical behavior as a positive interaction. Actually, this setting has the potential to yield sub-optimal performance, as each item makes a distinct contribution to the user's interest. For example, purchased items should be given more importance than clicked ones. Hence, we propose a general automatic sampling framework, named AutoSAM, to non-uniformly treat historical behaviors. Specifically, AutoSAM augments the standard sequential recommendation architecture with an additional sampler layer to adaptively learn the skew distribution of the raw input, and then sample informative sub-sets to build more generalizable SRS. To overcome the challenges of non-differentiable sampling actions and also introduce multiple decision factors for sampling, we further introduce a novel reinforcement learning based method to guide the training of the sampler. We theoretically design multi-objective sampling rewards including Future Prediction and Sequence Perplexity, and then optimize the whole framework in an end-to-end manner by combining the policy gradient. We conduct extensive experiments on benchmark recommender models and four real-world datasets. The experimental results demonstrate the effectiveness of the proposed approach. We will make our code publicly available after the acceptance.

TimeMAE: Self-Supervised Representations of Time Series with Decoupled Masked Autoencoders

Enhancing the expressive capacity of deep learning-based time series models with self-supervised pre-training has become ever-increasingly prevalent in time series classification. Even though numerous efforts have been devoted to developing self-supervised models for time series data, we argue that the current methods are not sufficient to learn optimal time series representations due to solely unidirectional encoding over sparse point-wise input units. In this work, we propose TimeMAE, a novel self-supervised paradigm for learning transferrable time series representations based on transformer networks. The distinct characteristics of the TimeMAE lie in processing each time series into a sequence of non-overlapping sub-series via window-slicing partitioning, followed by random masking strategies over the semantic units of localized sub-series. Such a simple yet effective setting can help us achieve the goal of killing three birds with one stone, i.e., (1) learning enriched contextual representations of time series with a bidirectional encoding scheme; (2) increasing the information density of basic semantic units; (3) efficiently encoding representations of time series using transformer networks. Nevertheless, it is a non-trivial to perform reconstructing task over such a novel formulated modeling paradigm. To solve the discrepancy issue incurred by newly injected masked embeddings, we design a decoupled autoencoder architecture, which learns the representations of visible (unmasked) positions and masked ones with two different encoder modules, respectively. Furthermore, we construct two types of informative targets to accomplish the corresponding pretext tasks. One is to create a tokenizer module that assigns a codeword to each masked region, allowing the masked codeword classification (MCC) task to be completed effectively...

FormerTime: Hierarchical Multi-Scale Representations for Multivariate Time Series Classification

Deep learning-based algorithms, e.g., convolutional networks, have significantly facilitated multivariate time series classification (MTSC) task. Nevertheless, they suffer from the limitation in modeling long-range dependence due to the nature of convolution operations. Recent advancements have shown the potential of transformers to capture long-range dependence. However, it would incur severe issues, such as fixed scale representations, temporal-invariant and quadratic time complexity, with transformers directly applicable to the MTSC task because of the distinct properties of time series data. To tackle these issues, we propose FormerTime, an hierarchical representation model for improving the classification capacity for the MTSC task. In the proposed FormerTime, we employ a hierarchical network architecture to perform multi-scale feature maps. Besides, a novel transformer encoder is further designed, in which an efficient temporal reduction attention layer and a well-informed contextual positional encoding generating strategy are developed. To sum up, FormerTime exhibits three aspects of merits: (1) learning hierarchical multi-scale representations from time series data, (2) inheriting the strength of both transformers and convolutional networks, and (3) tacking the efficiency challenges incurred by the self-attention mechanism. Extensive experiments performed on $10$ publicly available datasets from UEA archive verify the superiorities of the FormerTime compared to previous competitive baselines.

One Person, One Model--Learning Compound Router for Sequential Recommendation

Deep learning has brought significant breakthroughs in sequential recommendation (SR) for capturing dynamic user interests. A series of recent research revealed that models with more parameters usually achieve optimal performance for SR tasks, inevitably resulting in great challenges for deploying them in real systems. Following the simple assumption that light networks might already suffice for certain users, in this work, we propose CANet, a conceptually simple yet very scalable framework for assigning adaptive network architecture in an input-dependent manner to reduce unnecessary computation. The core idea of CANet is to route the input user behaviors with a light-weighted router module. Specifically, we first construct the routing space with various submodels parameterized in terms of multiple model dimensions such as the number of layers, hidden size and embedding size. To avoid extra storage overhead of the routing space, we employ a weight-slicing schema to maintain all the submodels in exactly one network. Furthermore, we leverage several solutions to solve the discrete optimization issues caused by the router module. Thanks to them, CANet could adaptively adjust its network architecture for each input in an end-to-end manner, in which the user preference can be effectively captured. To evaluate our work, we conduct extensive experiments on benchmark datasets. Experimental results show that CANet reduces computation by 55 ~ 65% while preserving the accuracy of the original model. Our codes are available at https://github.com/icantnamemyself/CANet.