Artificial Intelligence

Comp science proponents from the Binghamton University and the New York State University have developed a computation algorithm that foresees flight delays more accurately than the current methodologies in use. Sina Khanmohammadi, lead author of the project and PhD candidate at the Binghamton University, mentions that their initiative is without a doubt unique as it makes use of qualitative variables including weather and security risks together with the conventional numerical ones. With this take, their application has proved it can surpass traditional flight delay frameworks with regard to accuracy and speed (training time).

Flight delays are often predicted via computer models which are fed with backfill delay info from previous flights. These computation structures are integrated with artificial neural networks (ANN) which assess various variables to predict an outcome. These self-learning frameworks are trained to identify patterns by themselves. The addition of more variables in an ANN brings in more accuracy in terms of categorization. Historical data on the other hand slows down the ANN’s capability in detecting pattern configurations.

Reference

https://www.sciencedaily.com/releases/2016/11/161114103905.htm

Computer science specialists are viewing the current exponential data growth as a dire problem that needs urgent solutions; since its becoming more and more incomprehensible to humans. Huawei nonetheless intends to address the concern by manufacturing an advanced AI powered smartphone. Project specialists at the corporation assert that the device will transcend further from a personal assistant for daily activities to a human double/clone.

The gadget will be special due to its capabilities of proactively assessing the physical world in the same way the humans do. The gadget’s computer vision will serve its eyes, with a smart voice being the ears, the air sensor as the nose and the taste sensor as the tongue. Various robotic technologies will represent the body, with informed local/conscious decision-making frameworks serving as the mind. Combining these elements will facilitate the full replication of a human’s ability to understand the globe’s environs. It’s without doubt a compelling vision from one of the leading corporations in the mobile domain.

Reference

https://techcrunch.com/2016/10/11/huawei-puts-1m-into-a-new-ai-research-partnership-with-uc-berkeley/

Learners from the Brigham Young Univerisity have developed the VuePod, an immersive visualization environment entailing 12 high definition 55-inch 3D televisions, connected to a computer capable of supporting high-end, graphics-sophisticated gaming. Movement on the massive screen is controlled via a Wii remote which interacts with Kinnect-like Bluetooth device, as the 3D glasses create the dizzying added dimensions. The computer-powered mega TV is not nonetheless not meant for gaming, it’s for engineering purposes. It has already proved effective in enhancing the way engineers view environmental concerns; by allowing users to virtually float/hover/wander through 3D environs that are otherwise difficult to access. The images are created from point data recorded on an aircraft’s LIDAR, which scans the landscape and records millions of topographical references. The above mentioned data type is also attained from combined photographs taken from low-cost drones.

A dataset currently available for investigation at the VuePod is a canyon zone beneath a south Idaho plateau. With 3D glasses and a Wii controller, one can drop into the canyon from above, and navigate from one end to the other.

Reference

https://www.sciencedaily.com/releases/2014/12/141219160600.htm

M

A research study presented at the European Respiratory Society’s International Congress was the first of its kind to investigate how machine learning could aid/enhance the accuracy of lung disease diagnosis.

The above methodology entailed gathering of lung function data from 968 people, via a series of operations/checks including the spirometry test, which assesses breathing rates with regard to air volume and flow; followed by a body plethysmography which examines the static lung space and air resistance, and finally a diffusion test which ascertains the amount of oxygen and other gases transversing into the lung’s air sacs. Analysis of results is based on expert ideals/views plus the American Thoracic Society’s guidelines, which aim at developing patterns out of the findings.

Further, the research team incorporated an algorithm series to the lung parameters, coupled with clinical variables including age and body mass index. This approach enables the application to generate an accurate suggestion for the most likely effective diagnosis. The advancement also integrates clinicians’ complex reasoning employed in determining treatment procedures, but with a more objective and standardized approach to in turn ensure any bias is negated.

The researchers aim at testing the algorithm in different communities to enhance the system’s decision capability and transcend it to a globally validated clinical diagnosis approach.

Reference

https://www.sciencedaily.com/releases/2016/09/160904181255.htm

The price fluctuation of fine wines can now be accurately predicted via a novel artificial intelligence program methodology developed by researchers at the University College London. The approach could prove useful to fine wine investors, enabling them to make more informed decisions about their portfolio management, and subsequently increase the net trade of wine. The project team members intend to employ a similar approach to trading of other alternative assets such as classic cars.

Being the first time AI has been integrated in the world of fine wine, the team used two forms of machine learning, namely the Gaussian process regression and the more complicated ‘multi-task feature learning’, which was first unveiled by UCL scientists in 2006 but has overseen tremendous advancement recently. The techniques were able to obtain the most salient info from a variety of sources, as opposed to their more standard counterparts which classify every data point as spurious, of interest or otherwise.

Reference

https://www.sciencedaily.com/releases/2015/08/150804202706.html

Scientists at Uppsala University have utilized algorithmic modelling in a research project aimed at comprehending how brain tumors arise. Published in the journal of EBioMedicine, the study demonstrates how health proponents could employ large scale data to uncover with new disease mechanisms while identifying new treatment targets.

The application has assessed large amounts of patient data, afterwards providing a hypothesis on “what causes what” in cancer cells. According to the computer model, mesenchymal glioblastoma (an aggressive brain tumor type), can be partly attributed to alterations in the Annexin A2 gene. A follow up undertaking tested the inhibition of Annexin A2 features in patient cancer cells, and subsequently reported that cancer cells were either nullified or they transformed into a less aggressive form. The findings assert that data analysis could be employed to predict genes/proteins that influence the development of tumors. Furthermore, the SICS based method has been tested with other cancer forms displaying promising results, though additional enquiry is requisite to fine-tune the settings.

Reference

https://www.sciencedaily.com/releases/2016/09/160919093919.htm