Negative Sampling Method Research Articles

Aggregating neighborhood information for negative sampling for knowledge graph embedding

Knowledge graphs, as linked data, can be extracted from texts in triple form that illustrate the structure of “entity–relation–entity.” Knowledge graph embedding (KGE) models are used to map entities and relations into a continuous vector space with semantic constraints so as to learn a knowledge graph with fact triples. In the KGE model training process, both positive and negative triples are necessarily provided. Thus, negative sampling methods are meaningful in generating negative samples based on the representations of entities and relations. This paper proposes an innovative neighborhood knowledge selective adversarial network (NKSGAN), which leverages the representation of aggregating neighborhood information to generate high-quality negative samples for enhancing the performances of the discriminator. Experiments are conducted on widely used standard datasets such as FB15k, FB15k-237, WN18 and WN18RR to evaluate our model for link prediction task. The results present the superiority of our proposed NKSGAN than other baseline methods, indicating that the negative sampling process in NKSGAN is effective in generating high-quality negative samples for boosting KGE models.

Neglog: Homology-Based Negative Data Sampling Method for Genome-Scale Reconstruction of Human Protein-Protein Interaction Networks.

Rapid reconstruction of genome-scale protein–protein interaction (PPI) networks is instrumental in understanding the cellular processes and disease pathogenesis and drug reactions. However, lack of experimentally verified negative data (i.e., pairs of proteins that do not interact) is still a major issue that needs to be properly addressed in computational modeling. In this study, we take advantage of the very limited experimentally verified negative data from Negatome to infer more negative data for computational modeling. We assume that the paralogs or orthologs of two non-interacting proteins also do not interact with high probability. We coin an assumption as “Neglog” this assumption is to some extent supported by paralogous/orthologous structure conservation. To reduce the risk of bias toward the negative data from Negatome, we combine Neglog with less biased random sampling according to a certain ratio to construct training data. L2-regularized logistic regression is used as the base classifier to counteract noise and train on a large dataset. Computational results show that the proposed Neglog method outperforms pure random sampling method with sound biological interpretability. In addition, we find that independent test on negative data is indispensable for bias control, which is usually neglected by existing studies. Lastly, we use the Neglog method to validate the PPIs in STRING, which are supported by gene ontology (GO) enrichment analyses.

Medical Concept Representation Learning from Multi-source Data.

Representing words as low dimensional vectors is very useful in many natural language processing tasks. This idea has been extended to medical domain where medical codes listed in medical claims are represented as vectors to facilitate exploratory analysis and predictive modeling. However, depending on a type of a medical provider, medical claims can use medical codes from different ontologies or from a combination of ontologies, which complicates learning of the representations. To be able to properly utilize such multi-source medical claim data, we propose an approach that represents medical codes from different ontologies in the same vector space. We first modify the Pointwise Mutual Information (PMI) measure of similarity between the codes. We then develop a new negative sampling method for word2vec model that implicitly factorizes the modified PMI matrix. The new approach was evaluated on the code cross-reference problem, which aims at identifying similar codes across different ontologies. In our experiments, we evaluated cross-referencing between ICD-9 and CPT medical code ontologies. Our results indicate that vector representations of codes learned by the proposed approach provide superior cross-referencing when compared to several existing approaches.

WalkRanker: A Unified Pairwise Ranking Model With Multiple Relations for Item Recommendation

Top-N item recommendation techniques, e.g., pairwise models, learn the rank of users' preferred items through separating items into positive samples if user-item interactions exist, and negative samples otherwise. This separation results in an important issue: the extreme imbalance between positive and negative samples, because the number of items with user actions is much less than those without actions. The problem is even worse for "cold-start" users. In addition, existing learning models only consider the observed user-item proximity, while neglecting other useful relations, such as the unobserved but potentially helpful user-item relations, and high-order proximity in user-user, item-item relations. In this paper, we aim at incorporating multiple types of user-item relations into a unified pairwise ranking model towards approximately optimizing ranking metrics mean average precision (MAP), and mean reciprocal rank (MRR). Instead of taking statical separation of positive and negative sets, we employ a random walk approach to dynamically draw positive samples from short random walk sequences, and a rank-aware negative sampling method to draw negative samples for efficiently learning the proposed pairwise ranking model. The proposed method is compared with several state-of-the-art baselines on two large and sparse datasets. Experimental results show that our proposed model outperforms the other baselines with average 4% at different top-N metrics, in particular for cold-start users with 6% on average.

Effects of short-term warming and nitrogen addition on the quantity and quality of dissolved organic matter in a subtropical Cunninghamia lanceolata plantation.

Increasing temperature and nitrogen (N) deposition are two large-scale changes projected to occur over the coming decades. The effects of these changes on dissolved organic matter (DOM) are largely unknown. This study aimed to assess the effects of warming and N addition on the quantity and quality of DOM from a subtropical Cunninghamia lanceolata plantation. Between 2014 and 2016, soil solutions were collected from 0–15, 15–30, and 30–60 cm depths by using a negative pressure sampling method. The quantity and quality of DOM were measured under six different treatments. The spectra showed that the DOM of the forest soil solution mainly consisted of aromatic protein-like components, microbial degradation products, and negligible amounts of humic-like substances. Warming, N addition, and warming + N addition significantly inhibited the concentration of dissolved organic carbon (DOC) in the surface (0–15 cm) soil solution. Our results suggested that warming reduced the amount of DOM originating from microbes. The decrease in protein and carboxylic acid contents was mostly attributed to the reduction of DOC following N addition. The warming + N addition treatment showed an interactive effect rather than an additive effect. Thus, short-term warming and warming + N addition decreased the quantity of DOM and facilitated the migration of nutrients to deeper soils. Further, N addition increased the complexity of the DOM structure. Hence, the loss of soil nutrients and the rational application of N need to be considered in order to prevent the accumulation of N compounds in soil.

Effects of nitrogen deposition on the concentration and spectral characteristics of dissolved organic matter in soil solution in a young Cunninghamia lanceolata plantation.

To study the effects of nitrogen deposition on the concentration and spectral characteristics of dissolved organic matter (DOM) in the forest soil solution from the subtropical Cunninghamia lanceolata plantation, using negative pressure sampling method, the dynamics of DOM in soil solutions from 0-15 and 15-30 cm soil layer was monitored for two years and the spectroscopic features of DOM were analyzed. The results showed that nitrogen deposition significantly reduced the concentration of dissolved organic carbon (DOC), and increased the aromatic index (AI) and the humic index (HIX), but had no significant effect on dissolved organic nitrogen (DON) concentration in both soil layers. There was obvious seasonal variation in DOM concentration of the soil solution, which was prominently higher in summer and autumn than in spring and winter.Fourier-transform infrared (FTIR) absorption spectrometry indicated that the DOM in forest soil solution had absorption peaks in the similar position of six regions, being the highest in wave number of 1145-1149 cm-1. Three-dimensional fluorescence spectra indicated that DOM was mainly consisted of protein-like substances (Ex/Em=230 nm/300 nm) and microbial degradation products (Ex/Em=275 nm/300 nm). The availability of protein-like substances from 0-15 cm soil layer was reduced in the nitrogen treatments. Nitrogen deposition significantly reduced the concentration of DOC in soil solution, maybe largely by reducing soil pH, inhibiting soil carbon mineralization and stimulating plant growth. In particular, the decline of DOC concentration in the surface layer was due to the production inhibition of the protein-like substances and carboxylic acids. Short-term nitrogen deposition might be beneficial to the maintenance of soil fertility, while the long-term accumulation of nitrogen deposition might lead to the hard utilization of soil nutrients.

Effects of forest regeneration patterns on the quantity and chemical structure of soil solution dissolved organic matter in a subtropical forest.

Using the negative pressure sampling method, the concentrations and spectral characte-ristics of dissolved organic matter (DOM) of soil solution were studied at 0-15, 15-30, 30-60 cm layers in Castanopsis carlesii forest (BF), human-assisted naturally regenerated C. carlesii forest (RF), C. carlesii plantation (CP) in evergreen broad-leaved forests in Sanming City, Fujian Pro-vince. The results showed that the overall trend of dissolved organic carbon (DOC) concentrations in soil solution was RF>CP>BF, and the concentration of dissolved organic nitrogen (DON) was highest in C. carlesii plantation. The concentrations of DOC and DON in surface soil (0-15 cm) were all significantly higher than in the subsurface (30-60 cm). The aromatic index (AI) was in the order of RF>CP>BF, and as a whole, the highest AI was observed in the surface soil. Higher fluorescence intensity and a short wave absorption peak (320 nm) were observed in C. carlesii plantation, suggesting the surface soil of C. carlesii plantation was rich in decomposed substance content, while the degree of humification was lower. A medium wave absorption peak (380 nm) was observed in human-assisted naturally regenerated C. carlesii forest, indicating the degree of humification was higher which would contribute to the storage of soil fertility. In addition, DOM characte-ristics in 30-60 cm soil solution were almost unaffected by forest regeneration patterns.

A novel one-class SVM based negative data sampling method for reconstructing proteome-wide HTLV-human protein interaction networks.

Protein-protein interaction (PPI) prediction is generally treated as a problem of binary classification wherein negative data sampling is still an open problem to be addressed. The commonly used random sampling is prone to yield less representative negative data with considerable false negatives. Meanwhile rational constraints are seldom exerted on model selection to reduce the risk of false positive predictions for most of the existing computational methods. In this work, we propose a novel negative data sampling method based on one-class SVM (support vector machine, SVM) to predict proteome-wide protein interactions between HTLV retrovirus and Homo sapiens, wherein one-class SVM is used to choose reliable and representative negative data, and two-class SVM is used to yield proteome-wide outcomes as predictive feedback for rational model selection. Computational results suggest that one-class SVM is more suited to be used as negative data sampling method than two-class PPI predictor, and the predictive feedback constrained model selection helps to yield a rational predictive model that reduces the risk of false positive predictions. Some predictions have been validated by the recent literature. Lastly, gene ontology based clustering of the predicted PPI networks is conducted to provide valuable cues for the pathogenesis of HTLV retrovirus.

Mixing a Grounded Theory Approach with a Randomized Controlled Trial Related to Intimate Partner Violence: What Challenges Arise for Mixed Methods Research?

Little is known about how to systematically integrate complex qualitative studies within the context of randomized controlled trials. A two-phase sequential explanatory mixed methods study was conducted in Canada to understand how women decide to disclose intimate partner violence in emergency department settings. Mixing a RCT (with a subanalysis of data) with a grounded theory approach required methodological modifications to maintain the overall rigour of this mixed methods study. Modifications were made to the following areas of the grounded theory approach to support the overall integrity of the mixed methods study design: recruitment of participants, maximum variation and negative case sampling, data collection, and analysis methods. Recommendations for future studies include: (1) planning at the outset to incorporate a qualitative approach with a RCT and to determine logical points during the RCT to integrate the qualitative component and (2) consideration for the time needed to carry out a RCT and a grounded theory approach, especially to support recruitment, data collection, and analysis. Data mixing strategies should be considered during early stages of the study, so that appropriate measures can be developed and used in the RCT to support initial coding structures and data analysis needs of the grounded theory phase.