PyTorchKR Papers

업데이트된 논문

총 2편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2503.22054v2econ.EM공식 코드 · Papers with Code

tempdisagg: A Python Framework for Temporal Disaggregation of Time Series Data

tempdisagg is an open-source Python package that estimates a high-frequency time series from low-frequency benchmarks and, optionally, high-frequency indicators, such that the estimates aggregate exactly to the benchmarks. It implements the regression-based estimators of Chow-Lin, Fern\'andez and Litterman, the dynamic Chow-Lin model, the Denton and Denton-Cholette benchmarking procedures and uniform distribution, with the formulas and defaults of the R package tempdisagg, together with the autoregressive distributed-lag model of Proietti as an experimental extension. All generalised least squares estimators share one engine in which the residual covariance is applied through its triangular factor, so that one likelihood evaluation costs $O(n\,n_l^2)$ instead of $O(n^2 n_l)$. The package reports complete inference (coefficients, standard errors, likelihood-based criteria, prediction variances and a full profile of the autoregressive parameter with identification diagnostics), accepts additional exact linear constraints, multiple indicators and aggregation weights, and provides exact non-negative and cross-sectional reconciliation, rolling-origin backtesting, revision analysis, conformity diagnostics and a common-factor test. Agreement with R is enforced by an automated test suite over 29 reference cases: with a fixed autoregressive parameter predictions and statistics agree to relative differences between $10^{-15}$ and $10^{-7}$, and with an estimated parameter they agree to about $10^{-8}$ once the difference in optimiser tolerance is removed. Experiments with series whose high-frequency values are known illustrate how the reported tools support the choice of method.

arXiv:2410.23160v3cs.LG공식 코드 · Papers with Code

FlexTSF: A Flexible Forecasting Model for Time Series with Variable Regularities

Generalizing across time series with irregular temporal structures poses a persistent challenge for universal time series forecasting models. Existing approaches often assume regular sampling or depend heavily on imputation, which limits their applicability in real-world scenarios where irregularities are prevalent due to diverse sensing devices and recording practices. To address this gap, we propose FlexTSF, a flexible and domain-adaptive forecasting model specifically designed for time series with variable temporal regularities. At its foundation lies the IVP Patcher, a continuous-time patching module leveraging Initial Value Problems (IVPs) to inherently handle uneven sampling intervals, variable sequence lengths, and missing values. Built upon a decoder-only Transformer architecture, FlexTSF integrates normalized timestamp inputs and domain-specific statistical descriptors through a specialized causal self-attention mechanism, enabling adaptability across domains. Extensive experiments on 16 datasets demonstrate FlexTSF's effectiveness, significantly outperforming existing models in classic forecasting scenarios, zero-shot generalization, and low-resource fine-tuning conditions.