MotivationIn this work, researchers explored a simple extension of the current Transformer Networks with gating, named Gated Transformer Networks (GTN) for the multivariate time series classification

MotivationIn this paper, a transformer-based framework for unsupervised learning of multivariate time series was proposed for the first time. MethodThe overall pipeline is as follow: As we can see

MotivationFirst, the long-range dependencies of the time-series sequences are not well captured. Second, the interactions of multiple variables are generally not represented in features. To address th

MotivationThere exists highly variable lengths and sparse labeling in practice, to tackle the issues, a kind of unsupervised scalable representation was proposed this paper. Method Backbone module use

MethodIt’s the first iteration of MTSC datasets, which contains 30 datasets with equal series length per problem. The MTSC datasets of UEA archive consist of three types of sources, they are HAR(Human

MotivationA new synthetic series genaration method was proposed, and the average time series based on the extention of DBA(Dynamic Barycenter Averaging) was as a synthetic sample. MethodsThe paper f

IntroductionThe paper mainly introcude the solution for optimizing the RAG system, and the evaluation framework&tools are also included, which can be seen as below: More specifically, three p

MotivationA new efficient finetuning method based on quantile quantization was proposed, which realized the goal finetuning a 65B model on a single 48GB GPU, preserving full 16-bit task performance at

MotivationFinetuning full-parameter model is expensive. MethodUsing the low-rank A&B to represent the finetuned parameters, the detailed pipeline is as follow: ResultsThere are really abundant

MotivationIn the context of NLU tasks, the prior work reveals that prompt-tuning doesn’t perform well for normal-sized pretrained model and hard sequence labeling tasks(that means difficult tasks, lik