2023年2月1日 · MICN: Multi-scale Local and Global Context Modeling for Long-term Series Forecasting Huiqiang Wang , Jian Peng , Feihu Huang , Jince Wang , Junhui Chen , Yifei Xiao Published: 01 Feb 2023, Last Modified: 15 Feb 2023 ICLR 2023 …

termed as Multi-scale Isometric Convolution Network (MICN), is more efficient with linear complexity with respect to the sequence length. Our experiments on five benchmark datasets show that compared with state-of-the-art methods, MICN yields 18.2% and 24.5 relative improvements for multivariate and univariate time series, respectively.

%PDF-1.5 %¿÷¢þ 2183 0 obj /Linearized 1 /L 1587429 /H [ 3215 456 ] /O 2187 /E 140479 /N 22 /T 1574056 >> endobj 2184 0 obj /Type /XRef /Length 143 /Filter ...

as Multi-scale Isometric Convolution Network (MICN), is more efficient with linear complexity about the sequence length with suitable convolution kernels. Our experiments on six benchmark datasets show that compared with state-of-the-art methods, MICN yields 17.2% and 21.6% relative improvements for multivariate and univariate time series ...

Enter your feedback below and we'll get back to you as soon as possible. To submit a bug report or feature request, you can use the official OpenReview GitHub repository: Report an issue

2024年9月25日 · To address these limitations, we propose a novel framework called Med-MICN (Medical Multi-dimensional Interpretable Concept Network). Med-MICN provides interpretability alignment for various angles, including neural symbolic reasoning, concept semantics, and saliency maps, which are superior to current interpretable methods.

PatchTST (Nie et al., 2022); (3) Models use Convolutional Neural Networks (CNNs), like MICN and TimesNet (Wang et al., 2022; Wu et al., 2022). It’s worth noting that these categories are not rigidly isolated but rather represent key approaches within the broader landscape of time series analysis. To illustrate this point, let’s delve into two

(2022) and MICN Wang et al. (2023) proposed multi-scale hybrid decomposition approach based on Moving Average to extract various seasonal and trend-cyclical parts of time series. However, the real-world time series are usually intricate which are influenced by multiple factors and can be hardly disentangled.

regression. MICN (Wang et al., 2023) also decomposes input series into seasonal and trend terms, and then integrates the global and local context for forecasting. As for the multi-periodicity analysis, N-BEATS (Oreshkin et al., 2019) fits the time …

Interest-based item representation generated by MICN shared with original model takes user diverse interest information in whole model. The contributions of this paper can be summarized as follows: • A new framework to learn interest-based item representations by introducing user Multi Interests Capsule Network(MICN).