Field-Programmable Gate Array (FPGA) accelerators have proven successful in
handling latency- and resource-critical deep neural network (DNN) inference
tasks. Among the most computationally intensive operations in a neural network
(NN) is the dot product between the feature and weight vectors. Thus, some
previous FPGA acceleration works have proposed mapping neurons with quantized
inputs and outputs directly to lookup tables (LUTs) for hardware
implementation. In these works, the boundaries of the neurons coincide with the
boundaries of the LUTs. We propose relaxing these boundaries and mapping entire
sub-networks to a single LUT. As the sub-networks are absorbed within the LUT,
the NN topology and precision within a partition do not affect the size of the
lookup tables generated. Therefore, we utilize fully connected layers with
floating-point precision inside each partition, which benefit from being
universal function approximators, with rigid sparsity and quantization enforced
only between partitions, where the NN topology becomes exposed to the circuit
topology. Although cheap to implement, this approach can lead to very deep NNs,
and so to tackle challenges like vanishing gradients, we also introduce skip
connections inside the partitions. The resulting methodology can be seen as
training DNNs with a specific sparsity pattern that allows them to be mapped to
much shallower circuit-level networks, thereby significantly improving latency.
We validate our proposed method on a known latency-critical task, jet
substructure tagging, and on the classical computer vision task, the digit
classification using MNIST. Our approach allows for greater function
expressivity within the LUTs compared to existing work, leading to lower
latency NNs for the same accuracy.

利用全连接层内的浮点精度作为每个分区内的刚性稀疏性和量化，我们提出了将整个子网络映射到单个查找表（LUT）中的方法，并通过引入跳跃连接来解决梯度消失等挑战，从而显着提高延迟。

NeuraLUT：在布尔可综合函数中隐藏神经网络密度

NeuraLUT: Hiding Neural Network Density in Boolean Synthesizable  Functions

Sequence classification has a wide range of real-world applications in
different domains, such as genome classification in health and anomaly
detection in business. However, the lack of explicit features in sequence data
makes it difficult for machine learning models. While Neural Network (NN)
models address this with learning features automatically, they are limited to
capturing adjacent structural connections and ignore global, higher-order
information between the sequences. To address these challenges in the sequence
classification problems, we propose a novel Hypergraph Attention Network model,
namely Seq-HyGAN. To capture the complex structural similarity between sequence
data, we first create a hypergraph where the sequences are depicted as
hyperedges and subsequences extracted from sequences are depicted as nodes.
Additionally, we introduce an attention-based Hypergraph Neural Network model
that utilizes a two-level attention mechanism. This model generates a sequence
representation as a hyperedge while simultaneously learning the crucial
subsequences for each sequence. We conduct extensive experiments on four data
sets to assess and compare our model with several state-of-the-art methods.
Experimental results demonstrate that our proposed Seq-HyGAN model can
effectively classify sequence data and significantly outperform the baselines.
We also conduct case studies to investigate the contribution of each module in
Seq-HyGAN.

提出了一种名为 Seq-HyGAN 的超图注意力网络模型用于解决序列分类的挑战，并使用两级注意机制生成序列表示为超边，同时学习每个序列的关键子序列。实验结果表明，Seq-HyGAN 模型可以有效地分类序列数据并显着优于基线。