arXiv subjects

Takumi Fujimoto

Publications and source records attributed to Takumi Fujimoto.

3 recordsLinked to original sources

arXiv2026

EPOC: Endpoint-Preserving Online Correction With Compressed Residual State for Multi-Horizon Time Series Forecasting

Completed multi-horizon forecasts provide residual feedback for a fixed forecaster, but retaining full residual blocks increases auxiliary state. We propose Endpoint-Preserving Online Correction (EPOC) with a compressed residual state. It stores low-order discrete cosine transform (DCT) coefficients and the final value of the preceding residual block. Within each channel, the endpoint is shared across component-wise online ridge regressions that also use current-forecast coefficients. The fitted DCT correction is blended with the base forecast. We evaluate eight multivariate series with DLinear and PatchTST, three seeds, and two training variants, yielding 96 matched fixed-base conditions at a 24-step horizon. EPOC achieves mean condition-wise reductions in mean squared error (MSE) and mean absolute error (MAE) of 15.40% and 9.35% from the uncorrected base, respectively, with a median of 6,352 B in retained auxiliary arrays. It has lower paired MSE than the $δ$-Adapter, COSA, FAC, and OMPB in a majority of conditions and uses less state than each. Full ELF achieves the largest mean MSE reduction, 19.29%, but its median retained state is 474,048 B ($\times$75 relative to EPOC). Equal-size summary controls favor the endpoint by 1.65--2.20% in paired MSE; a coefficient-reconstructed endpoint yields similar accuracy to the observed endpoint, highlighting its role as a shared input. Increasing the retained DCT component count from 4 to 8 adds 1.00 percentage point of MSE reduction for 5,728 B. On jointly trained bases, EPOC lowers MSE by 16.69--20.15% relative to globally blended TEFL-style adapters applied to the same base. The code and numerical records are available at https://github.com/keiotakmin/endpoint-preserving-residual-correction.

cs.LG↗

arXiv2026

When Does Online Adaptation Pay on the Edge? A Leakage-Free Evaluation of Warmup, Learning-Rate Selection, and Resource Trade-offs for Time-Series Forecasting

Online adaptation can help edge time-series forecasting under distribution drift, but its measured benefit is sensitive to evaluation choices. We study six public multivariate streams, including building-sensor and smart-meter data, under a leakage-free streaming protocol. We identify two additional sources of comparison bias. First, the warmup budget of the static baseline has a two-sided effect: insufficient warmup undertrains the baseline, whereas excessive warmup can degrade its pre-drift generalization. Across six dataset-backbone settings, the estimated adaptation benefit changes by 3.0 to 18.8 percentage points (pp) over the 1,000-20,000-step warmup range. Second, comparing SGD with momentum (SGD+m) and Adam at a shared default learning rate conflates optimizer quality with rate sensitivity. We select both the warmup budget and each optimizer's online rate using a held-out pre-drift validation slice without accessing test data. Under this validation-only procedure, Adam outperforms SGD+m in 310 of 360 evaluated cells, while 4 Adam cells remain below the static baseline. We further characterize accuracy against adaptation-state memory and A100-measured per-update latency for full, head-only, and calibration-based adaptation. In the evaluated PatchTST frontier settings, several parameter-efficient variants are nondominated on the adaptation-state-memory axis. Smart-meter analyses also show that reported gains depend on meter-selection rules. These findings support a validation-only commissioning procedure, while target-device latency and energy remain to be measured. Code, data, and all reported numbers: https://github.com/keiotakmin/tsf-edge-adaptation.

cs.LG↗

arXiv2025

eagle: early approximated gradient based learning rate estimator

We propose EAGLE update rule, a novel optimization method that accelerates loss convergence during the early stages of training by leveraging both current and previous step parameter and gradient values. The update algorithm estimates optimal parameters by computing the changes in parameters and gradients between consecutive training steps and leveraging the local curvature of the loss landscape derived from these changes. However, this update rule has potential instability, and to address that, we introduce an adaptive switching mechanism that dynamically selects between Adam and EAGLE update rules to enhance training stability. Experiments on standard benchmark datasets demonstrate that EAGLE optimizer, which combines this novel update rule with the switching mechanism achieves rapid training loss convergence with fewer epochs, compared to conventional optimization methods.

cs.LG↗