Modern applications are built from third-party libraries like a wall from bricks. Their usage is the only option to complete the project in a reasonable time, spending a sensible budget, so it's a usual practice. However, taking all the bricks indiscriminately may not be such a good idea. If there are several options, it is useful to take time to analyze open libraries in order to choose the best one.

With ML projects still on the rise we are yet to see integrated solutions in almost every device around us. The need for processing power, memory and experimentation has led to machine learning and DL frameworks targeting desktop computers first. However once trained, a model may be executed in a more constrained environment on a smartphone or on an IoT device. A particularly interesting environment to run the model on is browser. Browser-based solutions may be used on a wide range of devices, desktop and mobile, online and offline. The topic of this post is how to prepare a model for the in-browser usage.

This post presents an end-to-end implementations of a model creation in Python and Node.js. The end goal is to create a model and to use it in a browser. I'll use TensorFlow and TensorFlow.js as main frameworks. One could train a model in Python and convert it to JS. Alternative is to train a model directly in javascript, hence omitting the conversion step.

I have more experience in Python and use it in my everyday work. I occasionally use javascript, but have very little experience in the contemporary front-end development. My hope from this post that python developers with little JS experience could use it to kick start their JS usage.

Code reviews are definitely necessary and useful. It's a way to impart knowledge, educate, control a task, improve code quality and formatting, fix bugs. Moreover, you can notice high-level errors related to the architecture and algorithms used. So it's a must-have practice, except that people get tired quickly. Therefore, static analysis perfectly complements reviews and helps to detect a variety of inconspicuous errors and typos. Let's look at a decent example on this topic.

This is the concluding part of my article devoted to a statistical analysis of police shootings and criminality among the white and the black population of the United States. In the first part, we talked about the research background, goals, assumptions, and source data; in the second part, we investigated the national use-of-force and crime data and tracked their connection with race.

Author: Sergey Lukyanchikov, Sales Engineer at InterSystems

Challenges of real-time AI/ML computations

We will start from the examples that we faced as Data Science practice at InterSystems:

• A “high-load” customer portal is integrated with an online recommendation system. The plan is to reconfigure promo campaigns at the level of the entire retail network (we will assume that instead of a “flat” promo campaign master there will be used a “segment-tactic” matrix). What will happen to the recommender mechanisms? What will happen to data feeds and updates into the recommender mechanisms (the volume of input data having increased 25000 times)? What will happen to recommendation rule generation setup (the need to reduce 1000 times the recommendation rule filtering threshold due to a thousandfold increase of the volume and “assortment” of the rules generated)?
• An equipment health monitoring system uses “manual” data sample feeds. Now it is connected to a SCADA system that transmits thousands of process parameter readings each second. What will happen to the monitoring system (will it be able to handle equipment health monitoring on a second-by-second basis)? What will happen once the input data receives a new bloc of several hundreds of columns with data sensor readings recently implemented in the SCADA system (will it be necessary, and for how long, to shut down the monitoring system to integrate the new sensor data in the analysis)?
• A complex of AI/ML mechanisms (recommendation, monitoring, forecasting) depend on each other’s results. How many man-hours will it take every month to adapt those AI/ML mechanisms’ functioning to changes in the input data? What is the overall “delay” in supporting business decision making by the AI/ML mechanisms (the refresh frequency of supporting information against the feed frequency of new input data)?

In this article, I’ll try to describe roughly how I’m creating a device, from the idea to realization of the usable prototype.

My name is Dmitriy Dudarev. I develop electronics, and really enjoy creating different portable devices. I also enjoy music. Half a year ago, I borrowed acoustic guitar from my friend, so that I could learn to play it from lessons on Youtube and tablatures. It was a hard work. Maybe, I did something wrong, or tried not so hard, or fine motor skills prevented multiplication in my predecessors’ community. In any case, I couldn’t achieve anything but sounds of raspy strings. My resentment was strengthened by constant string detune. And others didn’t really enjoy much listening my uneven Nothing else matters for thousand times.
But, going through all these tortures, I haven’t forgotten the main rule of electronics engineer. If something exists, you can put microcontroller there. Or, at least, make a portable electronic modification.

In the previous part of this article, I talked about the research background, goals, assumptions, source data, and used tools. Today, without further ado, let's say together…

Do the police in the US really shoot black people more often than white people? Is use of lethal force connected with race? How is crime related to race? What are the odds of getting shot by the police if you are white and if you are black? We're taking public data and python with pandas to shed some light on these questions, propaganda and politics set far aside.

We are proud to announce that we have built from ground up and released our high-quality (i.e. on par with premium Google models) speech-to-text Models for the following languages:

• English;
• German;
• Spanish;

You can find all of our models in our repository together with examples, quality and performance benchmarks. Also we invested some time into making our models as accessible as possible — you can try our examples as well as PyTorch, ONNX, TensorFlow checkpoints. You can also load our model via TorchHub.

About a year ago, we released Token2 Molto-1, the world's first programmable multi-profile hardware token. While Molto-1 is still the only solution of its kind currently available on the market, we will be soon releasing a new variation of a multi-profile hardware token, in a different form-factor and with a different set of features available.

While Molto-1 has its advantages, there were some shortcomings that we wanted to address, for example, it can only hold up to ten TOTP profiles, which is not enough for many users. Also, using NFC to program the device does not look very convenient for some users. There were also requests to have a backlight for the screen of the token, so it can be used in the dark. With Molto-2 we tried to address this and a few other concerns. So, we hereby present our new device model, Token2 Molto-2 with the following specifications:

TOKEN2 MOLTO-2 multi-profile programmable TOTP hardware token:

▣ RFC 6238 compliant

▣ supports up to 50 accounts/profiles

▣ USB-programmable with a Windows app

▣ RTC battery life: 8 years

▣ LCD screen battery: 3-4 months (rechargeable)

The table below shows the comparison between Molto-1 and Molto-2

I offer a solution to one beautiful task — writing code that outputs its text is valid for interpreters and compilers of different languages and is correctly executed when reversing its sources.

Not so long ago I learned about code that can be both interpreted in PHP and compiled to Java: PhpJava.java. As it turned out, this idea is not new: code which is valid for several compilers or interpreters is called a polyglot. It is possible to write such code because of the peculiarities of processing strings and comments in different interpreters or compilers.

In July 2020, the European Court of Justice invalidated an exchange of the personal data between the European Union and the United States. The times of the Safe Harbor and the Privacy Shield are over. Now what?

The National Internet Segment Reliability Research explains how the outage of a single Autonomous System might affect the connectivity of the impacted region with the rest of the world. Most of the time, the most critical AS in the region is the dominant ISP on the market, but not always.

As the number of alternate routes between AS’s increases (and do not forget that the Internet stands for “interconnected network” — and each network is an AS), so does the fault-tolerance and stability of the Internet across the globe. Although some paths are from the beginning more important than others, establishing as many alternate routes as possible is the only viable way to ensure an adequately robust network.

The global connectivity of any given AS, regardless of whether it is an international giant or regional player, depends on the quantity and quality of its path to Tier-1 ISPs.

Usually, Tier-1 implies an international company offering global IP transit service over connections with other Tier-1 providers. Nevertheless, there is no guarantee that such connectivity will be maintained all the time. For many ISPs at all “tiers”, losing connection to just one Tier-1 peer would likely render them unreachable from some parts of the world.

Sales and operations go hand in hand if you wish success for your organization. Even a hint of compromise in any of the aspects could be the reason for the downfall for your business, which is a scenario that no one would welcome. Keeping such a situation at a distance is possible if you have a data-driven business strategy.

Before offering any options of cooperation to our new client, we conducted a free preliminary analysis of competitors' strategies, as well as a review of search demand. We managed to surprise the client already at this stage.

Real Competitors’ Websites Analysis Case

Space exploration was always fascinating, and recent developments have reignited the interest to the heights never seen since the last man stood on the Moon. People argue about Mars exploration and features of spaceships as their grandparents would’ve done if the internet existed fifty years ago. I’m an electronics engineer working in the aerospace industry, so I know a thing or two about the technical background of this stuff — and I see that these things aren’t common knowledge, and people often have significantly skewed ideas about the reasons behind many things and decisions. Namely, I’d love to speak of some misconceptions about radiation hardened integrated circuits and the means of protection from radiation-induced damage.

Introduction

Against the backdrop of the coronavirus pandemic, the demand for cloud gaming services has noticeably increased. These services provide computing power to launch video games and stream gameplay to user devices in real-time. The most obvious advantage of this gaming type is that gamers do not need to have high-end hardware. An inexpensive computer is enough to run the client, spending time in self-isolation while the remote server carries out all calculations.

NVIDIA GeForce NOW is one of these cloud-based game streaming services. According to Google Trends, worldwide search queries for GeForce NOW peaked in February 2020. This correlates with the beginning of quarantine restrictions in many Asian, European, and North and South American countries, as well as other world regions. At the same time in Russia, where the self-isolation regime began in March, we see a similar picture with a corresponding delay.

Given the high interest in GeForce NOW, we decided to explore this service from an information security standpoint.

Do you know AI can save us from a worldwide pandemic?

Yeah, it's true. Our global researchers have touted these two buzzing technologies can provide a substantial social benefit to this worldwide health crisis.

Before I begin, I would like to take this moment to say THANK YOU to all our COVID-19 Warriors standing on the frontline and working day and night for us. We can’t thank them enough. Our healthcare staff, police, scientist, security guards, and sweepers. Their contribution is overwhelming and commendable ?

Discovering a drug for any medicine demands the joint efforts of the world's brightest minds. The process is notoriously long, complicated, and expensive. And that's how health experts are involved in searching COVID-19 medicine. In the midst of such a crisis, artificial intelligence solutions are offering a new hope that a cure might appear faster with it.

«Nothing makes you feel stupid quite like reading a scientific journal article» — writes the TV presenter and molecular biologist Adam Ruben. In a way, he's right — many of us get lost in the often confusing language of peer-reviewed papers. But the situation does not have to be hopeless. A bit of effort on the readers' part can go a long way. We looked at the techniques actual scientists use to navigate academic content. And compiled them into this two-part guide (Part 1: Getting Better at Understanding Academic Papers).

QEMU is a rather well-known application for emulation. Static analysis can help developers of complex projects such as QEMU catch errors at early stages and generally improve quality and reliability of a project. In this article, we will check the source code of the QEMU application for potential vulnerabilities and errors using the PVS-Studio static analysis tool.