Traditional approaches to variational inference rely on parametric families of variational distributions, with the choice of family playing a critical role in determining the accuracy of the resulting posterior approximation. Simple mean-field families often lead to poor approximations, while rich families of distributions like normalizing flows can be difficult to optimize and usually do not incorporate the known structure of the target distribution due to their black-box nature. To expand the space of flexible variational families, we revisit Variational Rejection Sampling (VRS) [Grover et al., 2018], which combines a parametric proposal distribution with rejection sampling to define a rich non-parametric family of distributions that explicitly utilizes the known target distribution. By introducing a low-variance reparameterized gradient estimator for the parameters of the proposal distribution, we make VRS an attractive inference strategy for models with continuous latent variables. We argue theoretically and demonstrate empirically that the resulting method--Reparameterized Variational Rejection Sampling (RVRS)--offers an attractive trade-off between computational cost and inference fidelity. In experiments we show that our method performs well in practice and that it is well-suited for black-box inference, especially for models with local latent variables.

Sphero, a toy manufacturer known for making simple, programmable robots for kids, has launched a new spinoff business to develop AI and robotics for law enforcement, first responders, and other government agencies. The new entity is called Company Six (CO6) and will build on technology Sphero had previously developed through its Public Safety Division. The company hasn't announced any clients or new projects, but promises to focus on'lightweight, yet highly advanced robotic solution that provides critical awareness for those we depend on the most, including police, fire, EMT, military, and others with dangerous jobs.' Sphero's Paul Berberian, who previously served in the US Air Force, will step down from his role as CEO and take a new title as Chairman of both companies, according to a report in CNet. 'This is an opportunity to continue to bring revolutionary robotics technology to new markets to improve the lives of more people, our future leaders, and people with essential and sometimes dangerous job functions,' he said in a prepared statement. Sphero says the company has sold more than four million robots since it was founded in 2010.

Weather and atmospheric patterns are often persistent. The simplest weather forecasting method is the so-called persistence model, which assumes that the future state of a system will be similar (or equal) to the present state. Machine learning (ML) models are widely used in different weather forecasting applications, but they need to be compared to the persistence model to analyse whether they provide a competitive solution to the problem at hand. In this paper, we devise a new model for predicting low-visibility in airports using the concepts of mixture of experts. Visibility level is coded as two different ordered categorical variables: cloud height and runway visual height. The underlying system in this application is stagnant approximately in 90% of the cases, and standard ML models fail to improve on the performance of the persistence model. Because of this, instead of trying to simply beat the persistence model using ML, we use this persistence as a baseline and learn an ordinal neural network model that refines its results by focusing on learning weather fluctuations. The results show that the proposal outperforms persistence and other ordinal autoregressive models, especially for longer time horizon predictions and for the runway visual height variable.

Sphero is introducing a new robot today on Kickstarter. Called RVR, which can either be pronounced just like it's spelled or like "rover," the robot is a development platform designed to be a bridge between educational robots like Sphero and more sophisticated and expensive platforms like Misty. It's mostly affordable, very expandable, and comes from a company with a lot of experience making robots. For a US $199 Kickstarter pledge, this looks like a very solid little robot, carefully thought-out and designed to be rugged and durable. It comes with an ambient light sensor, RGB color sensor, 9-axis IMU, IR sensors, magnetic encoders for the wheels, and a bunch of LEDs.

Sphero's been amusing us with its collection of robotic balls, like its adorable BB-8, for eight years. But lately the company has been getting away from the toy aspect of its products and embracing its educational potential. It's had an app that can be used to program many of its current bots for a while now, but that's only for budding coders -- what do kids interested in hardware have to tinker with? Indeed, Sphero is about to release its first robot specifically made to be physically modded, called the RVR. RVR -- that's pronounced "Rover" -- is a big change for Sphero.