Daily Tech Digest - May 16, 2021

Scientist develops an image recognition algorithm that works 40% faster than analogs

Convolutional neural networks (CNNs), which include a sequence of convolutional layers, are widely used in computer vision. Each layer in a network has an input and an output. The digital description of the image goes to the input of the first layer and is converted into a different set of numbers at the output. The result goes to the input of the next layer and so on until the class label of the object in the image is predicted in the last layer. For example, this class can be a person, a cat, or a chair. For this, a CNN is trained on a set of images with a known class label. The greater the number and variability of the images of each class in the dataset are, the more accurate the trained network will be. ... The study's author, Professor Andrey Savchenko of the HSE Campus in Nizhny Novgorod, was able to speed up the work of a pre-trained convolutional neural network with arbitrary architecture, consisting of 90-780 layers in his experiments. The result was an increase in recognition speed of up to 40%, while controlling the loss in accuracy to no more than 0.5-1%. The scientist relied on statistical methods such as sequential analysis and multiple comparisons.


Comprehensive Guide To Dimensionality Reduction For Data Scientists

The approaches for Dimensionality Reduction can be roughly classified into two categories. The first one is to discard less-variance features. The second one is to transform all the features into a few high-variance features. We will have a few of the original features in the former approach that do not undergo any alterations. But in the later approach, we will not have any of the original features, rather, we will have a few mathematically transformed features. The former approach is straightforward. It measures the variance in each feature. It claims that a feature with minimal variance may not have any pattern in it. Therefore, it discards the features in the order of their variance from the lowest to the highest. Backward Feature Elimination, Forward Feature Construction, Low Variance Filter and Lasso Regression are the popular techniques that fall under this category. The later approach claims that even a less-important feature may have a small piece of valuable information. It does not agree with discarding features based on variance analysis.


How security reskilling and automation can mend the cybersecurity skills gap

To understand the high demand for cybersecurity skills, consider how much has changed in IT—especially in the last year. From a rapid increase in cloud migrations to a huge shift toward remote work, IT teams everywhere have been forced to adapt quickly to keep up with the changing needs of their organizations. However, the rapid expansion of technology and explosion of remote work has kept IT busy enough. They don’t have the capacity to adequately handle responsibilities ranging from regular security hygiene to the patching and forensics surrounding the latest zero-day threat. ... With the difficulty of recruiting, hiring, and onboarding new cybersecurity experts from a small talent pool, consider investing in retraining your workforce to organically grow needed cybersecurity skills. Besides avoiding a lengthy headhunting process, this also makes clear economic sense. According to the Harvard Business Review, it can cost six times as much to hire from the outside rather than build talent from within. In addition, focusing on retraining opens up career progression for your best employees—building their skills, morale, and loyalty to your organization.


The Flow System: Leadership for Solving Complex Problems

One of the most significant limitations in today’s leadership practices is the lack of development. Most leadership training is disguised as leader education. These training efforts also do not include time for emerging leaders to practice their newly learned leadership skills. Without practice and the freedom to fail during the developmental stages, it is nearly impossible for emerging leadership to master skill. Another problem with leadership development is that most programs deliver training to everyone the same way. Most leadership development programs were initially designed as “one-size-fits-all” training. In The Flow System, we make great efforts to design leadership and team development around the contextual setting. We view leadership as a collective construct, not an individual construct. We incorporate the team as the model of leadership, and individual team members as leaders using a shared leadership model. This collective becomes the organization’s leadership model, from the lower ranks up to the executive level. 


5 Practices To Give Great Code Review Feedback

The first thing to do is to have a very clear context about the PR. Sometimes we want to go fast; we think we already know what our colleague wanted to do, the best way to do it, and we just skim through the description. However, it is much better to take some time and read the title and description of the PR carefully, especially the latter because we could find all the assumptions that guided our colleague. We could find a more detailed description of the task and perhaps a good description of the main issue they faced when developing it. This could give us all the information we need to perform a constructive review, taking into consideration all the relevant aspects of it. ... When reviewing a piece of code, focus on the most important parts: the logic, the choices of data structure and algorithms, whether all the edge cases have been covered in the tests, etc. Many of the other syntax/formatting elements should be taken care of by a tool, such as a linter, a formatter, a spell checker. etc. There is no point in highlighting them in a comment. The same idea holds on how the documentation is written. There should be some conventions, and it is OK to tell the contributor if they are not following them.


Machine learning does not magically solve your problems

Looking at the neural network approach we see that some of the manual tasks are absorbed into the neural network. Specifically, feature engineering and selection are done internally by the neural network. On the flipside, we have to determine the network architecture (number of layers, interconnectedness, loss function, etc) and tune the hyperparamers of the network. In addition, many other tasks such as assessing the business problem still need to be done. As with TSfresh/Lasso, the neural network is an approach that works well in a specific situation, and is not a quick nor automated procedure. A good way to frame to change from regression to the neural network is that instead of solving the problem manually, we build a machine that solves the problem for us. Adding this layer of abstraction allows us to solve problems we never thought we could solve, but that still takes a lot of time and money to create. ... Machine learning has some magical and awe-inspiring applications, extending the range of applications we thought possible to be solved using a computer. However, the awesome potential of machine learning does not mean that it automatically solves our challenges.


Five Tips For Creating Design Systems

A product experience that delights is usually designed with persistent visuals and consistent interaction. Users want to feel comfortable knowing that no matter where they navigate, they won’t be surprised by what they find. Repetition, in the case of product design, is not boring, but welcome. Design systems create trust with users. Another benefit is the increased build velocity from design and engineering teams. As designers, we are tasked with solving problems. We want to create a simple understanding of how our users can accomplish tasks in a workflow. Of course, we are tempted at times to invent new patterns to solve design problems. We often forget, in the minutia of design iterations, that we’ve already solved a particular problem in a prior project or in another part of the current product. This inefficiency can lead to wasted time, especially if those existing patterns and components have not been documented. In a single-person design team, the negative effects may not be as visible, but one can imagine the exponential nature of a larger design team consistently duplicating existing work or creating new patterns that, ultimately, create an inconsistent user experience.


A Gentle Introduction to Multiple-Model Machine Learning

Typically, a single output value is predicted. Nevertheless, there are regression problems where multiple numeric values must be predicted for each input example. These problems are referred to as multiple-output regression problems. Models can be developed to predict all target values at once, although a multi-output regression problem is another example of a problem that can be naturally divided into subproblems. Like binary classification in the previous section, most techniques for regression predictive modeling were designed to predict a single value. Predicting multiple values can pose a problem and requires the modification of the technique. Some techniques cannot be reasonably modified for multiple values. One approach is to develop a separate regression model to predict each target value in a multi-output regression problem. Typically, the same algorithm type is used for each model. For example, a multi-output regression with three target values would involve fitting three models, one for each target.


Why is Business Intelligence (BI) important?

The term “data-driven decision-making” doesn’t fully encapsulate one of its important subtexts: People almost always mean fast decisions. This distinction matters because it’s one of the capabilities that modern BI tools and practices enable: Decision-making that keeps pace (or close enough to it) with the speed at which data is produced. “Data is now produced so fast and in such large volumes that it is impossible to analyze and use effectively when using traditional, manual methods such as spreadsheets, which are prone to human error,” says Darren Turner, head of BI at Air IT. “The advantage of BI is that it automatically analyzes data from various sources, all accurately presented in one easy-to-digest dashboard.” Sure, everyone talks about the importance of speed and agility across technology and business contexts. But that’s kind of the point: If you’re not doing it, your competitors almost certainly are. ... “In a marketplace where the volume of data is ever-increasing, the ability for it to be processed and translated into sound business decisions is essential for better understanding customer behavior and outperforming competitors.”


What Is NFT (Non Fungible Tokens)? What Does NFT Stand For?

The bulk of NFTs are stored on the Ethereum network.. Certain NFTs, which store additional information that allows them to function differently are also supported by the blockchain. Ethereum, like bitcoin and dogecoin, is a cryptocurrency, but the blockchain frequently accepts such non-fungible tokens (NFTs), which store additional information that enables them to function differently Person tokens that are part of the Ethereum network that have extra information are known as NFTs. The extra content is the most important feature, as it allows them to be displayed as art, music, video (and so on) in JPGs, MP3s, photographs, GIFs, and other formats. They can be bought and sold like any other medium of art because they have value – and their value is largely dictated by supply and demand, much like physical art. But that doesn’t suggest, in any way, that there is just one digital version of NFT art available to purchase. One can obviously replicate them, much like the art prints of originals are used, bought and sold, but they won’t be of the same value as the original one.



Quote for the day:

"It is not fair to ask of others what you are not willing to do yourself." -- Eleanor Roosevelt

No comments:

Post a Comment