Sequential recommendation aims to model users' evolving preferences based on their historical interactions. Recent advances leverage Transformer-based architectures to capture global dependencies, but existing methods often suffer from high computational overhead, primarily due to discontinuous memory access in temporal encoding and dense attention over long sequences. To address these limitations, we propose FuXi-γ, a novel sequential recommendation framework that improves both effectiveness and efficiency through principled architectural design. FuXi-γ adopts a decoder-only Transformer structure and introduces two key innovations: (1) An exponential-power temporal encoder that encodes relative temporal intervals using a tunable exponential decay function inspired by the Ebbinghaus forgetting curve. This encoder enables flexible modeling of both short-term and long-term preferences while maintaining high efficiency through continuous memory access and pure matrix operations. (2) A diagonal-sparse positional mechanism that prunes low-contribution attention blocks using a diagonal-sliding strategy guided by the persymmetry of Toeplitz matrix. Extensive experiments on four real-world datasets demonstrate that FuXi-γ achieves state-of-the-art performance in recommendation quality, while accelerating training by up to 4.74× and inference by up to 6.18×, making it a practical and scalable solution for long-sequence recommendation. Code: https://github.com/Yeedzhi/FuXi-gamma.

Efficient and effective modeling of feature interactions is key to large-scale Click-Through Rate (CTR) prediction. Although existing feature interaction methods have improved the model accuracy, their computational consumption still increase exponentially with the number of feature fields and become severe efficiency bottleneck in real-world industrial scenarios. To address the issues, we propose an E fficient and E ffective NET work for large-scale CTR prediction named EENet . EENet presents a new alternating stacking architecture of implicit and explicit interaction layers, and each implicit layer in EENet can reduce both local computational and parameter load remarkably. EENet also designs a unified explicit interaction operation which can only use simple matrix multiplication to capture field-wise patterns. Moreover, the order of multiplications in EENet is rearranged to further decrease the computational complexity from quadratic to linear with respect to the number of feature fields. EENet thus can support the high efficiency in real-practice industrial scenarios with hundreds of feature fields. A set of extensive experiments is performed on two public datasets and one industrial dataset for effectiveness evaluation, and five larger-scale synthetic datasets for efficiency evaluation. The results highlight that our EENet can significantly outperform the state-of-the-art models in terms of both efficiency and scalability, while also maintaining superior effectiveness. Compared with DCNv2 and FiBiNet, EENet achieves 8.06 \(\times\) and 36.72 \(\times\) efficiency improvements in training, and 2.02 \(\times\) and 48.88 \(\times\) improvements in inference, respectively. Our solution and source code are available at https://github.com/Yeedzhi/EENet .

Convolutional Neural Networks (CNNs) can benefit from the computational reductions provided by the Winograd minimal filtering algorithm and weight pruning. However, harnessing the potential of both methods simultaneously introduces complexity in designing pruning algorithms and accelerators. Prior studies aimed to establish regular sparsity patterns in the Winograd domain, but they were primarily suited for small tiles, with domain transformation dictating the sparsity ratio. The irregularities in data access and domain transformation pose challenges in accelerator design, especially for larger Winograd tiles. This paper introduces “Winols,” an innovative algorithm-hardware co-design strategy that emphasizes the strengths of the large-tiling Winograd algorithm. Through a spatial-to-Winograd relevance degree evaluation, we extensively explore domain transformation and propose a cross-domain pruning technique that retains sparsity across both spatial and Winograd domains. To compress pruned weight matrices, we invent a relative column encoding scheme. We further design an FPGA-based accelerator for CNN models with large Winograd tiles and sparse matrix-vector operations. Evaluations indicate our pruning method achieves up to 80% weight tile sparsity in the Winograd domain without compromising accuracy. Our Winols accelerator outperforms dense accelerator by a factor of 31.7× in inference latency. When compared with prevailing sparse Winograd accelerators, Winols reduces latency by an average of 10.9×, and improves DSP and energy efficiencies by over 5.6× and 5.7×, respectively. When compared with the CPU and GPU platform, Winols accelerator with tile size 8× 8 achieves 24.6× and 2.84× energy efficiency improvements, respectively.