The aim of knowledge graph (KG) completion is to extend an incomplete KG with missing triples. Popular approaches based on graph embeddings typically work by first representing the KG in a vector space, and then applying a predefined scoring function to the resulting vectors to complete the KG. These approaches work well in transductive settings, where predicted triples involve only constants seen during training; however, they are not applicable in inductive settings, where the KG on which the model was trained is extended with new constants or merged with other KGs. The use of Graph Neural Networks (GNNs) has recently been proposed as a way to overcome these limitations; however, existing approaches do not fully exploit the capabilities of GNNs and still rely on heuristics and adhoc scoring functions. In this paper, we propose a novel approach, where the KG is fully encoded into a GNN in a transparent way, and where the predicted triples can be read out directly from the last layer of the GNN without the need for additional components or scoring functions. Our experiments show that our model outperforms state-of-the-art approaches on inductive KG completion benchmarks.

The aim of knowledge graph (KG) completion is to extend an incomplete KG with missing triples. Popular approaches based on graph embeddings typically work by first representing the KG in a vector space, and then applying a predefined scoring function to the resulting vectors to complete the KG. These approaches work well in transductive settings, where predicted triples involve only constants seen during training; however, they are not applicable in inductive settings, where the KG on which the model was trained is extended with new constants or merged with other KGs. The use of Graph Neural Networks (GNNs) has recently been proposed as a way to overcome these limitations; however, existing approaches do not fully exploit the capabilities of GNNs and still rely on heuristics and ad-hoc scoring functions. In this paper, we propose a novel approach, where the KG is fully encoded into a GNN in a transparent way, and where the predicted triples can be read out directly from the last layer of the GNN without the need for additional components or scoring functions. Our experiments show that our model outperforms state-of-the-art approaches on inductive KG completion benchmarks.

The aim of knowledge graph (KG) completion is to extend an incomplete KG with missing triples. Popular approaches based on graph embeddings typically work by first representing the KG in a vector space, and then applying a predefined scoring function to the resulting vectors to complete the KG. These approaches work well in transductive settings, where predicted triples involve only constants seen during training; however, they are not applicable in inductive settings, where the KG on which the model was trained is extended with new constants or merged with other KGs. The use of Graph Neural Networks (GNNs) has recently been proposed as a way to overcome these limitations; however, existing approaches do not fully exploit the capabilities of GNNs and still rely on heuristics and ad-hoc scoring functions. In this paper, we propose a novel approach, where the KG is fully encoded into a GNN in a transparent way, and where the predicted triples can be read out directly from the last layer of the GNN without the need for additional components or scoring functions.

ANN ARBOR--With a shortage of new tuberculosis drugs in the pipeline, a software tool from the University of Michigan can predict how current drugs--including unlikely candidates--can be combined in new ways to create more effective treatments. "This could replace our traditional trial-and-error system for drug development that is comparatively slow and expensive," said Sriram Chandrasekaran, U-M assistant professor of biomedical engineering, who leads the research. Dubbed INDIGO, short for INferring Drug Interactions using chemoGenomics and Orthology, the software tool has shown that the potency of tuberculosis drugs can be amplified when they are teamed with antipsychotics or antimalarials. "This tool can accurately predict the activity of drug combinations, including synergy--where the activity of the combination is greater than the sum of the individual drugs," said Shuyi Ma, a research scientist at the University of Washington and a first author of the study. "It also accurately predicts antagonism between drugs, where the activity of the combination is lesser. In addition, it also identifies the genes that control these drug responses."

With a shortage of new tuberculosis drugs in the pipeline, a software tool from the University of Michigan can predict how current drugs, including unlikely candidates, can be combined in new ways to create more effective treatments – leading to the design of TB drug regimens. Sriram Chandrasekaran, U-M assistant professor of biomedical engineering, who leads the research, said: "This could replace our traditional trial-and-error system for drug development that is comparatively slow and expensive. Dubbed INDIGO, short for INferring Drug Interactions using chemoGenomics and Orthology, the software tool has shown that the potency of tuberculosis drugs can be amplified when they are teamed with antipsychotics or antimalarials. Shuyi Ma, a research scientist at the University of Washington and a first author of the study, said: "This tool can accurately predict the activity of drug combinations, including synergy, where the activity of the combination is greater than the sum of the individual drugs. "It also accurately predicts antagonism between drugs, where the activity of the combination is lesser. In addition, it also identifies the genes that control these drug responses."

Similar to the growth in the number of vehicles in an urban area, the number of aircraft and the passengers they ferry, are in a phase of constant growth. Globally, the number of aircraft is expected to double from the base year of 2015 up to 2035. Since there are only limited number of airports and limited amount of space in each airport, this implies that each aircraft movement on the ground needs to be efficiently handled for faster turnaround. Faced with the pressure of managing multiple cost heads, airlines are now outsourcing their airport ground handling and cargo management services to specialist companies, and focusing on their core competence. While growth trends in passenger volumes tend to follow macroeconomic fundamentals, the growth in aircraft turnarounds are more immune to such highs and lows.

When Geoffrey von Maltzahn was first pitching farmers to try out his startup's special seeds, he sometimes told them, half-acknowledging his own hyperbole, that "if we're right, you shouldn't just see results in the field, you should be able to see them from outer space." As the co-founder of a company called Indigo Ag, von Maltzahn was hawking a probiotic that he hoped would increase their crop yields dramatically. "I never thought we'd ever actually test that idea," he says. In the three years since Indigo began selling naturally occurring organisms such as bacteria and fungi, spray-coated onto seeds, the company has grown to become perhaps the most valuable agtech company in the world. Pitchbook, for example, estimates Indigo's value at $3.5 billion.

"Both the cost and time savings were immediately recognizable." Indigo, "the world's first cultural department store for book lovers," is an adored and well-respected brand. As part of Indigo's corporate identity, they take pride in an elevated employee experience. Repeatedly ranked among the top 10 in Top Retail Employer Brand lists, it is no surprise that Indigo is flooded with over 2200 online applications every single week. Indigo approached Ideal as their number of applications continued to climb.

You might say the folks at Ideal were victims of their own success. Co-founders Somen Mondal and Shaun Ricci's latest company grew from a challenge they faced in their first venture, inspection and safety compliance management programs. "We started the first company, Field ID, as two people working out of an attic, and grew to the point we were hiring two salespeople a week," Ideal CEO Mondal says. "We had all the classic HR nightmares: we would hire with bias and ignore résumé details. We had terrible recruiter efficiency and bad quality hires. It was so bad we knew that one of every two hires we made would likely be fired."