Population based training of neural networks | DeepMind

We introduce a new method for training neural networks which allows an experimenter to quickly choose the best set of hyperparameters and model for the task. This technique – known as Population Based Training – trains and optimises a pool of networks at the same time, allowing the optimal set-up to be quickly found. Crucially, this adds no computational overhead, can be done as quickly as traditional techniques and is easy to integrate into existing machine learning pipelines.

Source: Population based training of neural networks | DeepMind

Laptops Are Great. But Not During a Lecture or a Meeting. – The New York Times

But a growing body of evidence shows that over all, college students learn less when they use computers or tablets during lectures. They also tend to earn worse grades. The research is unequivocal: Laptops distract from learning, both for users and for those around them. It’s not much of a leap to expect that electronics also undermine learning in high school classrooms or that they hurt productivity in meetings in all kinds of workplaces.

Source: Laptops Are Great. But Not During a Lecture or a Meeting. – The New York Times

The rise and fall of Limux [LWN.net]

The LiMux (or Limux)initiative in Munich has been heralded as an example of both the good andbad in moving a public administration away from proprietary systems. FreeSoftware Foundation Europe (FSFE) President Matthias Kirschner reviewed thehistory of the initiative—and its recent apparent downfall—in a talk atOpen Source Summit Europe in Prague. He also looked at the broaderimplications of the project as well as asking some questions thatfree-software advocates should consider moving forward.

Source: The rise and fall of Limux [LWN.net]