Diffeq.jl v6.4: Full GPU ODEs, Neural ODEs with Batching on GPUs, and More

Diffeq.jl v6.4: Full GPU ODEs, Neural ODEs with Batching on GPUs, and More

The ability to use stiff ODE solvers on the GPU, with automated tooling for matrix-free Newton-Krylov, faster broadcast, better Jacobian re-use algorithms, memory use reduction, etc. Here’s a nice showcase of DifferentialEquations.jl: Neural ODE with batching on the GPU (without internal data transfers) with high order adaptive implicit ODE solvers for stiff equations using matrix-free Newton-Krylov via preconditioned GMRES and trained using checkpointed adjoint equations. Now not just the non-stiff ODE solvers but the stiff ODE solvers allow for the initial condition to be a GPUArray, with the internal methods not performing any indexing in order to allow for all computations to take place on the GPU without data transfers.

Source: juliadiffeq.org

Early efforts to build a thermometer

Early efforts to build a thermometer

Scientists in the early 1800s wrestled with the problem of measuring temperature, often using brandy in a glass tube. It was a time when they did not understand what caused heat, and the idea that temperature was a measure of the speed with which atoms move was flatly rejected by London’s Royal Society:

“Early efforts at measuring temperature were purely empirical. However, exhaustive comparisons in the 1840s by French scientist Henri Victor Regnault showed that an ‘air ther­mometer’ — which measures changes in pressure of dry air in a sealed container — was superior to both in its reproducibility and designs of air thermometer calibrated at the freezing and boiling points of water gave consistent estimates of temperatures.

Source: delanceyplace.com

The Global Attention Span Is Getting Shorter

The Global Attention Span Is Getting Shorter

As a result, the length of time that content remains popular — a rough measurement of the global attention span — is decreasing, according to a recent large-scale analysis published in Nature Communications. ”“The public interest is getting saturated quicker with one topic because there’s more content produced in the early stages of a trend,” says study co-author Philipp Lorenz-Spreen, who studies modern information systems at Max Planck Institute for Human Development in Germany. This means that we are becoming interested in trends more quickly, but are also losing interest in the same content more swiftly, says Sune Lehmann, another study co-author, who is a physicist and mathematician at the Technical University of Denmark.

Source: onezero.medium.com

Jeff Bezos unveils a lunar lander and BE-7 rocket engine

Jeff Bezos unveils a lunar lander and BE-7 rocket engine

WASHINGTON — Jeff Bezos, chairman of Amazon and founder of Blue Origin, unveiled his space company’s lunar lander for the first time on Thursday. The Blue Moon lander can bring 3.6 metric tons to the lunar surface, according to Bezos. “I love Vice President Pence’s 2024 lunar landing goal,” Bezos said, adding that Blue Origin can meet that timeline “because we started this three years ago.”

Source: www.cnbc.com

Uber prices blockbuster IPO conservatively to raise $8.1B

Uber prices blockbuster IPO conservatively to raise $8.1B

Uber’s IPO came against a backdrop of turbulent financial markets, fueled by the trade dispute between the United States and China, as well as the plunging share price of Lyft Inc, which is down 23 percent from its IPO price in late March. The IPO was oversubscribed, but Uber settled for a lower price to avoid a repeat of Lyft’s IPO in late March, which priced strongly then plunged in trade. Lyft’s stock slumped nearly 11 percent on Wednesday to a record closing low of $52.91, well below the $72 per share IPO price, after the ride-hailing company reported a $1.1 billion quarterly loss.

Source: www.reuters.com

Lockdown – The coming war on general-purpose computing

Lockdown – The coming war on general-purpose computing

Figuring out what recalcitrant programs were doing and routing around media defects were core skills for computer programmers, especially in the era of fragile floppy disks and the rough-and-ready early days of software development. Anti-copying strategies only became more fraught as networks spread; once we had bulletin boards, online services, USENET newsgroups and mailing lists, the expertise of people who figured out how to defeat these authentication systems could be packaged up in software as little crack files. It created laws that made it illegal to tell people how to extract secrets from unlocking programs, and it created laws that made it illegal to host copyrighted works or the secrets.

Source: boingboing.net

‘Build More Housing’ Is No Match for Inequality

‘Build More Housing’ Is No Match for Inequality

According to Storper and Rodríguez-Pose, the notion that an insufficient supply of housing is a main cause of urban economic problems is based on a number of faulty premises. For Storper and Rodríguez-Pose, the rising spatial inequality between cities and metro areas stems from different kinds of economies that distinguish different kinds of cities, not from differences in housing costs. “Planning deregulation and housing costs are neither going to solve the problem of areas lagging behind, nor are they likely to have an impact on the economic development of dynamic cities,” Rodríguez-Pose and Storper write.

Source: www.citylab.com

TLS 1.3 is going to save us all, and other why IoT is still insecure (2017)

TLS 1.3 is going to save us all, and other why IoT is still insecure (2017)

Traffic to an IoT device passes through many different networks: the transit provider from the application server, the content delivery network used to deliver device traffic, the ISP to the building where the device sits. TLS is used in two ways in IoT devices: First, TLS is used to encrypt data in transit. Here’s what the same POST looks like with the TLS handshake — this is with TLS 1.2:

Depending on distance between the device and the application server and the latency of the server, this can be hundreds of milliseconds added.

Source: blog.cloudflare.com

Why are 2D vector graphics so much harder than 3D?

Why are 2D vector graphics so much harder than 3D?

A quote from the PDF specification, Chapter 2.1, “Imaging Model”:

By the time the W3C wanted to develop a 2D graphics markup language for the web, Adobe championed the XML-based PGML , which had the PostScript graphics model front and center.Microsoft’s competing format, VML , was based on GDI, which as we know was based on PostScript. Adobe’s PostScript proposed a novel solution to this: an algorithm to “snap” these paths to the coarser grids that printers had. Only Adobe’s PostScript could interpret a Type 1 Font, and only Adobe’s Type 1 Fonts had the custom hinting technology allowing them to be visible at small sizes.

Source: blog.mecheye.net

TensorFlow Graphics: Computer Graphics Meets Deep Learning

TensorFlow Graphics: Computer Graphics Meets Deep Learning

Introducing TensorFlow Graphics: Computer Graphics Meets Deep Learning
Posted by Julien Valentin and Sofien Bouaziz
Github repository: last few years have seen a rise in novel differentiable graphics layers which can be inserted in neural network architectures. At a high level, a computer graphics pipeline requires 3D objects and their absolute positioning in the scene, a description of the material they are made of, lights and a camera. TensorFlow Graphics comes with two 3D convolution layers, and one 3D pooling layer, allowing for instance the training of networks to perform semantic part classification on meshes as illustrated below and demonstrated in this Colab notebook.

Source: medium.com