How does driving an electric car compare to driving a similar gasoline car in terms of total pollution damages? While some comments about the future of mobility may suggest that electric cars are already clearly much more environmental friendly, the answer is not at all clear cut and depends on a lot of factors. For example, how does one weight local pollutions that mainly create health damages against carbon dioxide emmisions that contribute to global warming? While the local pollution damages of a gasoline car mainly depend on where the car is driven, a most relevant factor for electric cars is which power plants generate the required electricity. Also outside temperatures play an important role by affecting the efficiency of electric cars.

A monthly roundup of news about Artificial Intelligence, Machine Learning and Data Science. This is an eclectic collection of interesting blog posts, software announcements and data applications from Microsoft and elsewhere that I’ve noted over the past month or so.

By Imtiaz Adam, DeepLearn007.

Inspired by David Schoch’s blog post,Traveling Beerdrinker Problem.Check out his blog, he has some amazing posts!

The idea of having a 360 degree view of the entire job seeking and matching landscape has always been a dream of any labour economist. Just imagine, a dataset of CVs and job seekers matched with job advertisements and openings! The potential of such a dataset to answer existing questions on the labour market is incredible. One could investigate market power between worker and firms, information asymmetry within the matching process, or find out new growth clusters and skills needed to support these areas. So it was slightly unfortunate that I was not able to get my hands on such a dataset during my time in the government (I believe only Linkedin could capture something close to what I described).

On November 7th, Uwe Friedrichsen and I gave our talk from the JAX conference 2018: Deep Learning – a Primer again at the W-JAX in Munich.

Weighted likelihood mixture modeling and model based clustering

Binary Ensemble Neural Network (BENN) Binary neural networks (BNN) have been studied extensively since they run dramatically faster at lower memory and power consumption than floating-point networks, thanks to the efficiency of bit operations. However, contemporary BNNs whose weights and activations are both single bits suffer from severe accuracy degradation. To understand why, we investigate the representation ability, speed and bias/variance of BNNs through extensive experiments. We conclude that the error of BNNs is predominantly caused by the intrinsic instability (training time) and non-robustness (train \& test time). Inspired by this investigation, we propose the Binary Ensemble Neural Network (BENN) which leverages ensemble methods to improve the performance of BNNs with limited efficiency cost. While ensemble techniques have been broadly believed to be only marginally helpful for strong classifiers such as deep neural networks, our analyses and experiments show that they are naturally a perfect fit to boost BNNs. We find that our BENN, which is faster and much more robust than state-of-the-art binary networks, can even surpass the accuracy of the full-precision floating number network with the same architecture. …

One of the most difficult parts of learning anything (including data science) is staying motivated. At first, we’re excited about our new studies, and we progress quickly through the basics. But over time, as the work gets more challenging and other problems and pressures arise in our lives, it gets easier and easier to fall off the wagon.