Category Archives: Neural Networks

Competitive Technology, Neural Networks

Growing Fields of Man-Made Brainpower: How Neural Networks Work

Leave a comment

Neural Networks as ‘Artificial’ Brain

Human brain capacities can be imitated by neural networks and can portray a registering procedure. According to Andy Chun from City College of Hong Kong’s Bureau of Software engineering, with the use of neural networks, we can try to mimic how nature works when it comes to learning the process of several things.

Chun added that neural systems show how data is passed in the middle of the neurons to “acquire knowledge and learn new things”. The human brain is for the most part comprised of neurons, which are associated with by inward wiring called ‘neurotransmitters’.

The Essence of Neural Networks

Chun added that there are neural networks’ researches about speech recognition and generation, recognizing faces and images. Neural systems are superior to most AI technology at taking care of issues that are “perceptual in nature”, like talking and seeing. Chun believes that neural systems exceed expectations at perceiving designs.

In any case, as neural systems learn individually, engineers need to simply compose a small amount of possibly long line of codes because it is given that it would take “a huge number of lines” of code to program more smart AI frameworks. Profound learning requires a gigantic measure of figuring power, with quicker processors and more mind-boggling neural structures. All things considered, profound learning is behind Tesla’s self-driving autos, which show themselves by watching film of human drivers from everywhere throughout the world, Chun says.

What’s more, AlphaGo, the AI technology program that as of late beat the best on the planet of the Go tabletop game, took in its moves “from observing all diversions at any point played by people on the web”, according to Chun.

Neural Networks’ Influence on the government

Later on, Chun predicts that administrations will have the capacity to learn distinctive things considerably more precisely about its citizens in living, work, and wellbeing propensities and in transportation needs. Furthermore, he added that forecast will be a key advantage from this profound learning.
Specifically, he distinguishes medical advantages like analyzing and learning wellbeing examples of people and be ready to perhaps offer projects to keep nationals from becoming ill in any case.

Chun further added that the greatest obstacles are security policies and regulations. When you have a self-sufficient AI, at that point governments need to make sense of how to appoint duty, if a defective self-driving auto gets into a crash, for instance. In any case, it is critical to take note of that as governments utilize resident information to learn and remove learning.

Developments in the use of Neural Networks

Assets to urge new businesses to concentrate on profound learning are given in Hong Kong, according to Chun. Additionally, Singapore’s Administration Innovation Organization has recognized profound learning as a key concentration for 2017.

In the interim, in China, engineers of Baidu search engine had built a chatbot to help specialists in noting patients’ inquiries and proposing treatment choices. Also, a group from Northeastern College in China has lately built up a neural system that can recognize the area of flawed flags in microgrids, which are littler lattices that are associated with the fundamental power framework yet, can work without it. In conclusion, neural networks could in future ‘spare’ the world in an unexpected way while tackling issues that influence every one of our lives such security and the possible impact of technology.

Competitive Technology, Neural Networks

Deep Neural Networks For More Effective Security Systems

Leave a comment

Airports’ security is as important as national security. People, not only within the country but across the globe are going in and out of the country. The US Department of Homeland Security with the help of Google launched a contest in creating computer algorithms that will enable airports to identify hidden items in the images produced by checkpoint body scanner.

Gathering Data Scientists and Their ideas

The government aims to enhance security through advanced screening technology at airports. Kaggle which is already owned by Google and a site that holds over a million data scientists operates the said contests. The government is giving $1.5 million dollars for the contests that will run for 6 months.

Anthony Goldbloom, founder, and chief executive of Kaggle revealed that the said contest is an initiative to develop a technology named ‘deep neural networks’. These are complex mathematical systems with the ability to learn a particular task through huge data analysis. For instance, a neural network can identify what a cat is after analyzing millions of cats’ images.

Utilizing Neural Network Technology

Neural network technology is already used by Google and Facebook in several tasks. These include translation of languages, recognition of verbal commands from smartphones and identifying human faces in online images. While building algorithms were used to identify symptoms of lung cancer in CT scans in the past, neural networks are now developed to work with automated systems and read more precise body scans. This technology will not only enhance security but will also make airports services faster.

When it comes to conducting a contest, several managers and administrators find it a good idea. John W. Halinski, a security consultant considers the said contest as ‘crowdsourcing’ idea that will gather skills from different data scientists. Meanwhile, John Fortune, a program manager working in the Department of Homeland Security believes that the contest will find many people with high problem-solving skills. Moreover, Homeland Security and other agencies are in the process of discovering how neural networks can be used at security checkpoints like in the airport.

Efficiency of Neural Networks

Proponents of neural networks believe that it can upgrade airport security due to its capacity to learn data in a short time. Recently, Homeland Security supplied over a thousand three-dimensional body scans. But the scans are not shared and are not used for the contest. Volunteers from Transportation Security Administration assisted the working team in creating data by walking through a set of test scanners done in New Jersey laboratory.

Data gathered were used for analysis. The result shows that neural networks are efficient in doing security-related tasks like identifying hidden items. On the other hand, experts say that the technology is not perfect. According to some research, law offenders can change items or displaced the system to fool the system run by neural networks. In this case, the image-recognition system powered by neural networks might fail to see some concealed items.

Nevertheless, the government has seen the potential of using this technology to help human screeners in maintain top airport security. In the near future, the government and other organizations hope that neural networks will create breakthroughs in the security system and related tasks.