Education

Anglia Ruskin University researchers develop mobile system which detects tuberculosis with 98.4% accuracy

Researchers from Anglia Ruskin University test different machine learning methods to classify digital images created using biosensors for the presence of tuberculosis antibodies. The goal was to create a system which could process and classify the images on a portable phone. They were able to achieve a 98.4% accuracy doing so using a random forest method.

Berkeley Lab releases reinforcement learning training platform for training autonomous vehicles on traffic regulation

The US Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) has developed a driving simulation platform using reinforcement learning, with the goal of using it to train autonomous vehicles how to regulate traffic congestion.

Boston Public Schools''s plan to reconfigure start times for high school students using an optimisation algorithm backfires

Boston Public Schools intended to leverage traditional AI methods of planning and analytics to reduce sleep deprivation of teenagers due to early school start times. Two MIT graduates were appointed to the task by officials, facing the trade off of minimising the number of school buses required, maximise the number of students starting school after 8am and increasing parental happiness and satisfaction. However, the proposed solution resulted in fury amongst parents as the updated schedule would affect middle and elementary students'' start times as well. Although BPS aimed to reduce inequities, with almost 85 percent of the district getting new start times, many black and brown families would be negatively affected. The proposed change by the algorithm was not implemented.

Cambridge researchers develop a system to analyse cancer research papers to discover previously unexplored molecular biology links

Researchers at the University of Cambridge introduce a literature-based discovery (LBD) system to identify intermittently linked associations for cancer research in published literature. The LION LBD system uses convolutional neural networks and natural language processing to go through annotated databases of published research and come up with potential relations based on users searches. Results indicate at least a third of the proposed relations are useful and viable.

China''s Ministry of Education grades students''s essays in one in four Chinese schools using deep learning

The deep learning system, whose actual status is still unclear, has been used in around 60,000 schools and 120 million students have even graded over the last 10 years. The system uses deep learning to compare essays with established teaching experience and has built its own body of knowledge. Some anecdotal suggestion that ''brilliant'' outliers may not be appropriately rewarded.

Clarivate Analytics aims to improve article peer review efficiency by adding natural language processing to its ScholarOne platform

Clarivate Analytics'' has teamed up with UNSILO to add natural language processing abilities to its journal peer review software, ScholarOne. This allows papers submitted for review to be automatically summarised and checked against other papers for plagiarism.

Clinicial researchers at Imperial College London and the University of Edinburgh develop machine learning based software that could speed up treatment of patients showing signs of stroke or dementia

Scientists at Imperial College London and the University of Edinburgh have created a new software to improve stroke and dementia diagnosis in brain scans. With the use of machine learning the program is able to identify one the commonest causes of stroke and dementia, small vessel disease, more accurately than current methods. The researchers state that this technology may aid clinicians at administering the best treatment to patients more quickly and predicting a person''s likelihood of developing dementia.

Columbia Business School researchers determine that unusual language in news stories can forecast future market stress through natural language processing

Columbia Business School researchers applied natural language processing techniques to news stories about corporations and assess the unusualness of the language to determine if investors utilise this information and the market adjust accordingly. They found a positive correlation but time lag between unusualness and story sentiment and market stress.

Cornell University researchers increase processing efficiency 300x by developing Adaboost algorithm to automate detection of elephant rumblings

Researchers develop an algorithm to automatically detect elephant rumblings from recorded sounds at certain jungle sites in central Africa. This allows a better understanding of the elephant population density at certain times which would be difficult to obtain through visual methods. The Adaboost algorithm used was an improvement in processing efficiency of over 300x compared to manual processing, and a recall rate of at least 70%.

Coursera launches employee learning and skill gap assessment tool for businesses using machine learning

Coursera has introduced a feature which allows its corporate customers to determine how their employees are doing with online learning courses as well as identify skill gaps compared with other companies. This is done using machine learning and provides companies with metrics for their employees'' skill gains while giving them the opportunity to recommend training modules.

Cray''s CosmoFlow can now predict cosmological parameters with unprecedented accuracy with a deep learning 3D convolutional neural network

Cray, the National Energy Research Scientific Computing Center at Lawrence Berkeley National Laboratory (NERSC), and Intel have developed CosmoFlow, a deep learning 3D convolutional neural network (CNN). The system is aimed at predicting cosmological parameters with great accuracy. Such parameters include density, matter density fluctuations and power law index of the density perturbation spectrum after inflation. The research team achieved the same level of accuracy as existing experiments in estimating the values of Ωm and σ8, and five times less error than previous measurements using deep learning in estimating the value of Ns.

DeepMind develops a highly accurate machine learning method for predicting protein structures

DeepMind has developed a machine learning method to predict the 3D structure of proteins based on their genetic makeup. According to the company, the AlphaFold system is the most accurate methodology currently in the world for the prediction.

Dreambox offers students larger and faster gains in achievement through AI adaptive learning according to a Harvard report

Harvard conducted a study on the use of DreamBox Learning software in the Howard County Public School System (HCPSS) and the Rocketship Education charter school network. The study aimed to measure the impact of the AI adaptive learning system on student achievement. The key founding highlighted that the majority of students did not engage with the software as much as it was recommended, the software''s use was driven by teachers and schools rather than student preferences, while students spending more time on the software and following its recommendations saw larger and faster gains in achievement. Although results for causal impact on student achievement is positive, the evidence is mixed.

Georgia State University improved on-time enrolment by 3.3 percentage points using a machine learning chatbot to guide students during the process

Georgia State University trialled a chatbot to address incoming students'' specific concerns and ensure they were on track with university onboarding. This reduced summertime "melt" - wherein students accepted to university do not end up enrolling in the fall -by 3.3 percentage points.

Guilford College improves Wi-Fi and Bluetooth wireless services using machine learning

Guilford College uses Mist Systems'' Learning WLAN to monitor its wireless network and services to proactively detect anomalies and bottlenecks. The platform utilises machine learning to map the network and monitor it to detect any variations and helps identify root causes immediately in case of issues.

Harvard university seismologists developed a deep learning algorithm to predict the location of aftershocks with an AUC of .849

A group of researchers have developed a deep learning solution to predict the location of aftershocks based on static stress. The model has improved AUC from .583 to .849 and precision from 3% to 6%.

Illinois Institutes''s Chicago-Kent College of Law researchers predict outcomes of US Supreme Court decisions with 70% accuracy using machine learning

Researchers build a machine learning model to predict whether US Supreme Court decisions will affirm or reject the status quo, both on a case-level and individual judge level, with the model achieving approximately 70% accuracy.

Imperial College London researchers aim at designing a device that will improve dialysis procedure for patients using machine learning

A team from Imperial College London has leveraged AI to design a device aiming at improving dialysis for patients. The team used machine learning algorithms to optimise the shape of an arterio-venous fistulae (AVF) based on computer modelling techniques from the aerospace industry. The project stems from the similarity between how unsteady currents in blood flows in veins during dialysis and how unsteady air pockets flow over a plane. The prototype device has successfully passed preliminary tests in pigs.

MIT AI Lab predicts Alzheimer''s disease before close family members with advanced motion detection and analysis with machine learning

MIT’s Computer Science and Artificial Intelligence Laboratory used unobtrusive motion detection algorithms designed to detect falls to patients they suspect have early onset of Alzheimer''s. They use low-power wireless to record sleep, breathing, gait and other behavioural patterns. Supervised machine learning algorithms analyse the motion to determine ''normal'' patterns versus Alzheimer''s related such as repeating behaviour or agitation, depression and sleep disturbances.

MIT Department of Mechanical Engineering developed smart power outlets that distinguish dangerous electrical from benign spikes with 99.95% accuracy with machine learning

MIT’s Department of Mechanical Engineering have designed a ''smart power outlet'' to help identify the difference between harmless electrical arcs and dangerous ones for safety reasons. The supervised machine learning model was fed electrical current data, labelled as ''good'' or ''bad''. The system monitored the electric flows helping it learn to identify unique devices which posed a potential threat from their currents. The researchers claim that after sufficient training, their ''smart power outlet'' was able to identify dangerous electrical arcs with 99.95% accuracy.

MIT Researchers design tool to optimise product designs according to optimal efficiencies and performance objectives

Researchers at Computer Science and Artificial Intelligence Laboratory (CSAIL) and Columbia University have developed a tool that suggests ways to optimise their design to achieve different optimal efficiencies. By analysing designs using computer-aided design (CAD) programs the software navigates tradeoffs between design parameters, such as height, length, and radius of a product. The enhanced designs can be modified to satisfy performance objectives such as weight, balance, and durability.

MIT researchers aim to make cancer treatments less toxic with machine learning

Researchers at MIT have developed a machine learning model aiming to make chemotherapy and radiotherapy dosing regimens less toxic for glioblastoma patients, without compromising their effectiveness. The technology analyses current treatment regimes and optimises treatment plans to offer patients the most drastic but also less frequent dosage.

MIT researchers develop system that can identify individuals with 96% accuracy for smart homes improvement

MIT scientists develop system that crowdsources data to speed up drug discovery using neural networks

Scientists at MIT have developed a system that lets pharmaceutical companies share their data to speed up drug discovery, while keeping them confidential. An artificial neural network (ANN), trained on 1.4 million drug-protein pairs that are known to both do and do not interact, identifies new drug-protein interactions. The system achieved that with 95% accuracy.