Graph neural networks (GNNs) have delivered remarkable results in various fields. However, the rapid increase in the scale of graph data has introduced significant performance bottlenecks for GNN inference. Both computational complexity and memory usage have risen dramatically, with memory becoming a critical limitation. Although graph sampling-based subgraph learning methods can help mitigate computational and memory demands, they come with drawbacks such as information loss and high redundant computation among subgraphs. This paper introduces an innovative processing paradgim for distributed graph learning that abstracts GNNs with a new set of programming interfaces and leverages Just-In-Time (JIT) compilation technology to its full potential. This paradigm enables GNNs to highly exploit the computational resources of distributed clusters by eliminating the drawbacks of subgraph learning methods, leading to a more efficient inference process. Our experimental results demonstrate that on industry-scale graphs of up to \textbf{500 million nodes and 22.4 billion edges}, our method can produce a performance boost of up to \textbf{27.4 times}.

Deep learning owes a lot of its success to automatic differentiation. Popular libraries such as TensorFlow and PyTorch keep track of gradients over neural network parameters during training with both comprising high-level APIs for implementing the commonly used neural network functionality for deep learning. JAX is NumPy on the CPU, GPU, and TPU, with great automatic differentiation for high-performance machine learning research. Along with a Deep Learning framework, JAX has created a super polished linear algebra library with automatic differentiation and XLA support. JAX is a new machine learning library from Google designed for high-performance numerical computing.

Since the first principles of Knowledge Compilation (KC), most of the work    has been focused in finding a good compilation target language in terms of    compromises between compactness and expressiveness. The central idea    remained unchanged in the last fifteen years: an off-line, very hard, stage,    allows to ``compile'' the initial theory in order to guarantee    (theoretically) an efficient on-line stage, on a set of predefined queries    and operations.  We propose a new ``Just-in-Time'' approach    for KC. Here, any Knowledge Base (KB) will be immediately available for    queries, and the effort spent on past queries will be partly amortized for    future ones.  To guarantee efficient answers, we rely on the    tremendous progresses made in the practical solving    of SAT and incremental SAT applicative problems. Even if each query may    be theoretically hard, we  show that our approach outperforms    previous KC approaches on the set of classical problems used in the field,    and allows to handle problems that are out of the scope of current    approaches.