Core Concepts
IIP-Mixer proposes a novel MLP-based architecture for accurate battery Remaining Useful Life prediction by leveraging intra-inter patch mixing.
Abstract
Accurate estimation of Remaining Useful Life (RUL) crucial for battery management systems.
Attention-based networks like Transformers and Informer popular for time series forecasting.
IIP-Mixer uses MLPs for intra-inter patch mixing for RUL prediction.
Weighted loss function introduced for varying feature importance.
Experimental results show IIP-Mixer outperforms other time-series frameworks.
Stats
"Our experiments demonstrate that IIP-Mixer achieves competitive performance in battery RUL prediction, outperforming other popular time-series frameworks."
"The size of output O(i) intra is the same with input X(i), the process can be summarized as the following equations: O(i) intra = W 2σ(W 1X(i))"
"The size of output O(i) inter is the same with input X(i)T, the process can be summarized as the following equations: O(i) inter = W 4σ(W 3X(i)T)"
Quotes
"Our proposed model IIP-Mixer achieves the best experimental results."
"IIP-Mixer can capture the local and global temporal patterns in time series data."