Deep neural networks (DNNs) have recently been achieving state-of-the-art performance on a variety of pattern-recognition tasks, most notably visual classification problems. Given that DNNs are now able to classify objects in images with near-human-level performance, questions naturally arise as to what differences remain between computer and human vision. A recent study revealed that changing an image (e.g. of a lion) in a way imperceptible to humans can cause a DNN to label the image as something else entirely (e.g. mislabeling a lion a library). Here we show a related result: it is easy to produce images that are completely unrecognizable to humans, but that state-of-the-art DNNs believe to be recognizable objects with 99.99% confidence (e.g. labeling with certainty that white noise static is a lion). Specifically, we take convolutional neural networks trained to perform well on either the ImageNet or MNIST datasets and then find images with evolutionary algorithms or gradient ascent that DNNs label with high confidence as belonging to each dataset class. It is possible to produce images totally unrecognizable to human eyes that DNNs believe with near certainty are familiar objects, which we call “fooling images” (more generally, fooling examples). Our results shed light on interesting differences between human vision and current DNNs, and raise questions about the generality of DNN computer vision.
Anh Nguyen, Jason Yosinski, Jeff Clune
Source: [1412.1897] Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images
What browser caching does is “remember” the resources that the browser has already loaded. When a visitor goes to another page on your website your logo, CSS files, etc. do not need to be loaded again, because the browser has them “remembered” (saved). This is the reason that the first view of a web page takes longer than repeat visits.
Source: How to leverage browser caching of your website or blog
Wooden puzzles are great toys in their own right, but magnetize them, and watch the attraction grow! Because wooden puzzle pieces are thick, they adhere best to magnetic walls if they have a complete magnetic backing.
Making wooden puzzles magnetic takes a bit more time. First, trace each puzzle piece, keeping them in order to make best use of the magnetic sheet. Cut out the shapes and peel and stick. Because tracings tend to be a hair larger than the actual pieces, you will need to trim each piece.
Holding the craft knife at an angle, slide the blade around the puzzle piece, removing the soft magnetic backing which overlaps the edge.
Source: How To Make A Magnetic Puzzle | hubpages
Optimizing your images not only helps make your WordPress site load faster, but can also boost your SEO. This round-up of image optimization plugins includes tools for compression as well as SEO.
Source: 10 Best Image Optimization Plugins for Speeding Up WordPress
WordPress 4.8 is the first major release of 2017. Check out what is new in WordPress 4.8 with screenshots and new features that you should try.
Source: What’s New in WordPress 4.8
In this research note, we introduce congressbr, an R package for retrieving data from the Brazilian houses of legislature. The package contains easy-to-use functions that allow researchers to query the Application Programming Interfaces of the Chamber of Deputies and the Federal Senate, perform cleaning data operations, and store information in a format convenient for future analyses. We outline the main features of the package and demonstrate its use with some practical examples.
Source: SocArXiv Preprints | congressbr: An R Package for Analysing Data from Brazil’s Chamber of Deputies and Federal Senate