TSLANet: Rethinking Transformers for Time Series Representation Learning TSLANet is a novel lightweight convolutional model designed for diverse time series tasks. It introduces an Adaptive Spectral Block (ASB) that uses Fourier analysis to enhance feature representation and capture both long-term and short-term interactions while mitigating noise via adaptive thresholding. Additionally, an Interactive Convolution Block is introduced, leveraging self-supervised learning to refine TSLANet's capacity for decoding complex temporal patterns and improve robustness across different datasets. Comprehensive experiments show that TSLANet outperforms state-of-the-art models in classification, forecasting, and anomaly detection tasks, demonstrating resilience and adaptability across various noise levels and data sizes. The ASB employs Fourier analysis to transform time series data into the frequency domain, where adaptive thresholding is used to attenuate high-frequency noise. After processing, the inverse Fourier transform reconstructs the time-domain features, now with reduced noise and enhanced representations. The Interactive Convolution Block captures complex temporal patterns through convolutional operations. TSLANet also incorporates self-supervised pretraining to enhance model capabilities, especially on large datasets. TSLANet is lightweight and has a time complexity of O(N log N), demonstrating superior efficiency and speed compared to self-attention. It outperforms other models in various tasks, including classification, forecasting, and anomaly detection. The model's effectiveness is validated through extensive experiments on multiple datasets, showing its robustness and adaptability in handling diverse time series data. TSLANet's design balances local and global temporal feature extraction, making it a versatile solution for time series analysis.TSLANet: Rethinking Transformers for Time Series Representation Learning TSLANet is a novel lightweight convolutional model designed for diverse time series tasks. It introduces an Adaptive Spectral Block (ASB) that uses Fourier analysis to enhance feature representation and capture both long-term and short-term interactions while mitigating noise via adaptive thresholding. Additionally, an Interactive Convolution Block is introduced, leveraging self-supervised learning to refine TSLANet's capacity for decoding complex temporal patterns and improve robustness across different datasets. Comprehensive experiments show that TSLANet outperforms state-of-the-art models in classification, forecasting, and anomaly detection tasks, demonstrating resilience and adaptability across various noise levels and data sizes. The ASB employs Fourier analysis to transform time series data into the frequency domain, where adaptive thresholding is used to attenuate high-frequency noise. After processing, the inverse Fourier transform reconstructs the time-domain features, now with reduced noise and enhanced representations. The Interactive Convolution Block captures complex temporal patterns through convolutional operations. TSLANet also incorporates self-supervised pretraining to enhance model capabilities, especially on large datasets. TSLANet is lightweight and has a time complexity of O(N log N), demonstrating superior efficiency and speed compared to self-attention. It outperforms other models in various tasks, including classification, forecasting, and anomaly detection. The model's effectiveness is validated through extensive experiments on multiple datasets, showing its robustness and adaptability in handling diverse time series data. TSLANet's design balances local and global temporal feature extraction, making it a versatile solution for time series analysis.