Goto

Collaborating Authors

 group leader


Dual-Segment Clustering Strategy for Federated Learning in Heterogeneous Environments

Sun, Pengcheng, Liu, Erwu, Ni, Wei, Yu, Kanglei, Wang, Rui, Jamalipour, Abbas

arXiv.org Artificial Intelligence

Federated learning (FL) is a distributed machine learning paradigm with high efficiency and low communication load, only transmitting parameters or gradients of network. However, the non-independent and identically distributed (Non-IID) data characteristic has a negative impact on this paradigm. Furthermore, the heterogeneity of communication quality will significantly affect the accuracy of parameter transmission, causing a degradation in the performance of the FL system or even preventing its convergence. This letter proposes a dual-segment clustering (DSC) strategy, which first clusters the clients according to the heterogeneous communication conditions and then performs a second clustering by the sample size and label distribution, so as to solve the problem of data and communication heterogeneity. Experimental results show that the DSC strategy proposed in this letter can improve the convergence rate of FL, and has superiority on accuracy in a heterogeneous environment compared with the classical algorithm of cluster.


Group Leader for Data and Analytics Team

#artificialintelligence

Any convictions will be evaluated to determine if they directly relate to the responsibilities and requirements of the position. Having a conviction history will not automatically disqualify an applicant from being considered for employment.


Artificial intelligence can boost power, efficiency of even the best microscopes

#artificialintelligence

With the help of artificial intelligence, even already powerful microscopes can see better, faster and process more data. In a new study, published Friday in the journal Nature Methods, researchers used new machine learning algorithms to combine a pair of novel microscopy techniques. The marriage dramatically accelerated image processing and yielded crisp, accurate results. To capture speedy biological processes in 3D, like the beating heart of a fish larva, researchers rely on a method called light-field microscopy. The technique involves the collection of massive amounts data, and as a result, image processing can take days.


Robots are helping to advance developmental biology

#artificialintelligence

The study of developmental biology is getting a robotic helping hand. Scientists are using a custom robot to survey how mutations in regulatory regions of the genome affect animal development. These regions aren't genes, but rather stretches of DNA called enhancers that determine how genes are turned on and off during development. The team describes the findings--and the robot itself--on October 14 in the journal Nature. "The real star is this robot," says David Stern, a group leader at HHMI's Janelia Research Campus.


Online Knowledge Distillation with Diverse Peers

Chen, Defang, Mei, Jian-Ping, Wang, Can, Feng, Yan, Chen, Chun

arXiv.org Machine Learning

Distillation is an effective knowledge-transfer technique that uses predicted distributions of a powerful teacher model as soft targets to train a less-parameterized student model. A pre-trained high capacity teacher, however, is not always available. Recently proposed online variants use the aggregated intermediate predictions of multiple student models as targets to train each student model. Although group-derived targets give a good recipe for teacher-free distillation, group members are homogenized quickly with simple aggregation functions, leading to early saturated solutions. In this work, we propose Online Knowledge Distillation with Diverse peers (OKDDip), which performs two-level distillation during training with multiple auxiliary peers and one group leader. In the first-level distillation, each auxiliary peer holds an individual set of aggregation weights generated with an attention-based mechanism to derive its own targets from predictions of other auxiliary peers. Learning from distinct target distributions helps to boost peer diversity for effectiveness of group-based distillation. The second-level distillation is performed to transfer the knowledge in the ensemble of auxiliary peers further to the group leader, i.e., the model used for inference. Experimental results show that the proposed framework consistently gives better performance than state-of-the-art approaches without sacrificing training or inference complexity, demonstrating the effectiveness of the proposed two-level distillation framework.


Robots in Depth with Craig Schlenoff

Robohub

In this episode of Robots in Depth, Per Sjöborg speaks with Craig Schlenoff, Group Leader of the Cognition and Collaboration Systems Group and the Acting Group Leader of the Sensing and Perception Systems Group in the Intelligent Systems Division at the National Institute of Standards and Technology. They discuss ontologies and the significance of formalized knowledge for agile robotics systems that can quickly and even automatically adapt to new scenarios.


Taliban official: Group leader killed in drone strike

U.S. News

This photo taken by a freelance photographer Abdul Salam Khan using his smart phone on Sunday, May 22, 2016, purports to show the destroyed vehicle in which Mullah Mohammad Akhtar Mansour was traveling in the Ahmad Wal area in Baluchistan province of Pakistan, near Afghanistan's border. A senior commander of the Afghan Taliban confirmed on Sunday that the extremist group's leader, Mullah Mohammad Akhtar Mansour, has been killed in a U.S. drone strike.

  Country:
  Industry: Government > Military (1.00)

Taliban official: Group leader killed in drone strike

Associated Press

A senior commander of the Afghan Taliban confirmed on Sunday that the extremist group's leader, Mullah Akhtar Mansour, has been killed in a U.S. drone strike. Mullah Abdul Rauf, who recently reconciled with Mansour after initially rebelling against his ascension to the leadership, told The Associated Press that Mansour died in the strike late Friday "in the Afghanistan-Pakistan border area." The office of Afghan President Ashraf Ghani confirmed in a statement that the strike took place but could not confirm Mansour's death. Chief Executive Abdullah Abdullah, however, said that Mansour is "more than likely" dead. Speaking live on television as he chaired a Cabinet meeting, Abdullah said Mansour's death would have a positive impact on attempts to bring peace to Afghanistan, where the Taliban have been waging an insurgency for 15 years.


Taliban official: Group leader killed in drone strike

Associated Press

A senior commander with the Afghan Taliban says the militant group's leader Mullah Akhtar Mansour has been killed in a U.S. drone strike. Mullah Abdul Rauf told The Associated Press Sunday that Mansour died in the strike late Friday night. He says the strike took place "in the Afghanistan-Pakistan border area." The office of Afghan President Ashraf Ghani confirmed the strike but could not confirm Mansour's death. Chief Executive Abdullah Abdullah, however, says that Mansour is "more than likely" dead.