In recent years, review-based collaborative filtering (CF) has been extensively studied, which is an combination between natural language processing (NLP) and recommender systems. The core pattern behind CF is to first model user and item, and then adopts a relatively primitive interaction between them for personalized recommendation. This pattern is very similar to the issue of sequence matching in NLP, where sequence 1 and sequence 2 are matched with a fine-grained interaction leading to a better result. Therefore, there is a tremendous room for further improvement in current review-based CF to release the power of fine-grained interaction. To this end, we treat the user review set and item review set as two sequences, and design a multi-level matching attention layer for fine-grained interaction. In addition, we devise the aspect-level and review-level attention to measure the contribution of each review. Extensive experiments on 24 public datasets show that the proposed model consistently outperforms the state-of-theart approaches. More importantly, by selecting the relevant reviews according to the aspect attention score and review attention score, we can observe which specific item aspects that user mainly concerned and which item characteristic highly matched with the user preference, in which the recommendation interpretability can be enhanced.
White light interferometry is a well-established surface recovery
technique. In this paper, a white light signal processing algorithm
based on phase error compensation using spectrum selection is
proposed. The derived nonlinear phase distribution from the
correlogram is modeled as the combination of random errors and
systemic deviations. By developing a new, to the best of our
knowledge, recovery algorithm, the phase noise can be separated from
the linear map and significantly attenuated. Based on the proposed
algorithm, the spectrum features of white light LEDs and halogen lamps
are investigated in detail. The inner products defined by three
selected points are employed to generate a coefficient to evaluate the
linearity of an unwrapped phase map within a certain spectrum region.
The optimal spectrum range corresponding to the best measurement
performance can then be located where the coefficient approximates 1
and the spectrum energy stays relatively high. The simulations are
carried out under different levels of SNR and scan step noises, which
show that the new method can effectively reduce additional disturbance
from the recovered topography. In experiments, the system with the
proposed method is first calibrated by a step height standard (VLSI,
182.7
±
2.0
n
m
) with the repeatability of 0.44 nm. A
silicon wafer and three roughness standards are also tested to further
verify the robustness of the new method.
Keeping the generated fuzzy frequent itemsets up-to-date and discovering the new fuzzy frequent itemsets are challenging problems in dynamic databases. In this paper, the classical H-struct structure is extended to mining fuzzy frequent itemsets. The extended H-mine algorithm can use any t-norm operator to calculate the support of fuzzy itemset. The FP-tree-based structure called the Initial-FP-tree and the New-FP-tree are built to maintain the fuzzy frequent itemsets in the original database and the new inserted transactions respectively. The strategy of incremental mining of fuzzy frequent itemsets is achieved by breath-first-traversing the Initial-FP-tree and the New-FP-tree. All of the fuzzy frequent itemsets in the updated database can be obtained by traversing the Initial-FP-tree. The experiments on real datasets show that the proposed approach runs faster than the batch extended H-mine algorithm. Comparing with the existing algorithm for incremental mining fuzzy frequent itemsets, the proposed approach is superior in terms of the execution time. The memory cost of the proposed approach is lower than that of the existing algorithm when the minimum support threshold is low.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.