Recently, the embedding-based recommendation models (e.g., matrix factorization and deep models) have been prevalent in both academia industry due to their effectiveness flexibility. However, they also such intrinsic limitations as lacking explainability suffering from data sparsity. In this paper, we propose an end-to-end joint learning framework get around these without introducing any extra overhead by distilling structured knowledge a differentiable path-based model. Through extensive...

Just-in-time (JIT) compilation in dynamic programming languages can improve execution speed code with hot sections. However, that comes at the cost of increased memory usage for storage compiled and associated bookkeeping data, as well up-front time code.

In social networks, nodes (or users) interested in specific topics are often influenced by others. The influence is usually associated with a set of rather than single one. An interesting but challenging task for any given topic and node to find the that represents source or trigger thus identify those have greatest on as spreads. We it an NP-hard problem. This paper proposes effective framework deal this First, propagation represented Bayesian network. then construct model variant voter...

Just-in-time (JIT) compilation coupled with code caching are widely used to improve performance in dynamic programming language implementations. These caches, along the associated profiling data for hot code, however, consume significant amounts of memory. Furthermore, they incur extra JIT time their creation. On Android, current standard compiler and its caches not shared among processes—that is, runtime system maintains a private cache, data, each process. However, applications running on...

Real-world scenarios demand reasoning about process, more than final outcome prediction, to discover latent causal chains and better understand complex systems. It requires the learning algorithms offer both accurate predictions clear interpretations. We design a set of trajectory tasks on graphs with only source destination observed. present attention flow mechanism explicitly model leveraging relational inductive biases by basing our models graph networks. study way can effectively act...

Multi-type objects with multi-type relations are ubiquitous in real-world networks, e.g. bibliographic networks. Such networks also called heterogeneous information However, the research on clustering for is little. A new algorithm, NetClus, has been proposed recent two years. Although NetClus applied a network star schema, considering between center and all attribute linking to them, it ignores such as citation relations, which contain rich information. Hence, we think schema cannot be used...

In real-world scenarios, it is appealing to learn a model carrying out stochastic operations internally, known as computation graphs (SCGs), rather than learning deterministic mapping. However, standard backpropagation not applicable SCGs. We attempt address this issue from the angle of cost propagation, with local surrogate costs, called Q-functions, constructed and learned for each node in an SCG. Then, SCG can be trained based on these costs using backpropagation. propose entire framework...

The ability of reasoning beyond data fitting is substantial to deep learning systems in order make a leap forward towards artificial general intelligence. A lot efforts have been made model neural-based as an iterative decision-making process based on recurrent networks and reinforcement learning. Instead, inspired by the consciousness prior proposed Yoshua Bengio, we explore with notion attentive awareness from cognitive perspective, formulate it form message passing graphs, called neural...

Just-in-time (JIT) compilation coupled with code caching are widely used to improve performance in dynamic programming language implementations. These caches, along the associated profiling data for hot code, however, consume significant amounts of memory. Furthermore, they incur extra JIT time their creation. On Android, current standard compiler and its caches not shared among processes---that is, runtime system maintains a private cache, data, each process. However, applications running...

Planar transformer is one of the key technologies for miniaturization Electronic Power Conditioner used in Travelling-Wave Tube Amplifier. The parasitic parameters planar are difficult to model because nonlinear and multivariable characteristics. This paper presents a prediction based on binary particle swarm optimization extra-trees. has been trained verified by data set transformer. results confirm accuracy model.