Kai Zheng

Yuanyong Luo, Zhongxing Zhang, Richard Wu et al.

This preliminary white paper proposes a novel 8-bit floating-point data format HiFloat8 (abbreviated as HiF8) for deep learning. HiF8 features tapered precision. For normal value encoding, it provides 7 exponent values with 3-bit mantissa, 8 exponent values with 2-bit mantissa, and 16 exponent values with 1-bit mantissa. For denormal value encoding, it extends the dynamic range by 7 extra powers of 2, from 31 to 38 binades (notice that FP16 covers 40 binades). Meanwhile, HiF8 encodes all the special values except that positive zero and negative zero are represented by only one bit-pattern. Thanks to the better balance between precision and dynamic range, HiF8 can be simultaneously used in both forward and backward passes of AI training. In this paper, we will describe the definition and rounding methods of HiF8, as well as the tentative training and inference solutions. To demonstrate the efficacy of HiF8, massive simulation results on various neural networks, including traditional neural networks and large language models (LLMs), will also be presented.

Hung Vinh Tran, Tong Chen, Guanhua Ye et al.

Content-based Recommender Systems (CRSs) play a crucial role in shaping user experiences in e-commerce, online advertising, and personalized recommendations. However, due to the vast amount of categorical features, the embedding tables used in CRS models pose a significant storage bottleneck for real-world deployment, especially on resource-constrained devices. To address this problem, various embedding pruning methods have been proposed, but most existing ones require expensive retraining steps for each target parameter budget, leading to enormous computation costs. In reality, this computation cost is a major hurdle in real-world applications with diverse storage requirements, such as federated learning and streaming settings. In this paper, we propose Shapley Value-guided Embedding Reduction (Shaver) as our response. With Shaver, we view the problem from a cooperative game perspective, and quantify each embedding parameter's contribution with Shapley values to facilitate contribution-based parameter pruning. To address the inherently high computation costs of Shapley values, we propose an efficient and unbiased method to estimate Shapley values of a CRS's embedding parameters. Moreover, in the pruning stage, we put forward a field-aware codebook to mitigate the information loss in the traditional zero-out treatment. Through extensive experiments on three real-world datasets, Shaver has demonstrated competitive performance with lightweight recommendation models across various parameter budgets. The source code is available at https://github.com/chenxing1999/shaver

Yue Cui, Jiandong Xie, Kai Zheng

Long sequence time-series forecasting (LSTF) has become increasingly popular for its wide range of applications. Though superior models have been proposed to enhance the prediction effectiveness and efficiency, it is reckless to neglect or underestimate one of the most natural and basic temporal properties of time-series. In this paper, we introduce a new baseline for LSTF, the historical inertia (HI), which refers to the most recent historical data-points in the input time series. We experimentally evaluate the power of historical inertia on four public real-word datasets. The results demonstrate that up to 82\% relative improvement over state-of-the-art works can be achieved even by adopting HI directly as output.