Clinical electroencephalographic (EEG) data varies significantly depending on a number of operational conditions (e.g., the type and placement of electrodes, the type of electrical grounding used). This investigation explores the statistical differences present in two different referential montages: Linked Ear (LE) and Averaged Reference (AR). Each of these accounts for approximately 45% of the data in the TUH EEG Corpus. In this study, we explore the impact this variability has on machine learning performance. We compare the statistical properties of features generated using these two montages, and explore the impact of performance on our standard Hidden Markov Model (HMM) based classification system. We show that a system trained on LE data significantly outperforms one trained only on AR data (77.2% vs. 61.4%). We also demonstrate that performance of a system trained on both data sets is somewhat compromised (71.4% vs. 77.2%). A statistical analysis of the data suggests that mean, variance and channel normalization should be considered. However, cepstral mean subtraction failed to produce an improvement in performance, suggesting that the impact of these statistical differences is subtler.

Recurrent Neural Networks (RNNs) with sophisticated units that implement a gating mechanism have emerged as powerful technique for modeling sequential signals such as speech or electroencephalography (EEG). The latter is the focus on this paper. A significant big data resource, known as the TUH EEG Corpus (TUEEG), has recently become available for EEG research, creating a unique opportunity to evaluate these recurrent units on the task of seizure detection. In this study, we compare two types of recurrent units: long short-term memory units (LSTM) and gated recurrent units (GRU). These are evaluated using a state of the art hybrid architecture that integrates Convolutional Neural Networks (CNNs) with RNNs. We also investigate a variety of initialization methods and show that initialization is crucial since poorly initialized networks cannot be trained. Furthermore, we explore regularization of these convolutional gated recurrent networks to address the problem of overfitting. Our experiments revealed that convolutional LSTM networks can achieve significantly better performance than convolutional GRU networks. The convolutional LSTM architecture with proper initialization and regularization delivers 30% sensitivity at 6 false alarms per 24 hours.

The past year has been a momentous one for science and technology. From the detection of gravitational waves (predicted almost a century ago by Einstein) to the rise of virtual currencies like Bitcoin to the creation of genetically modified human embryos, 2017 was marked by all sorts of remarkable discoveries and innovations. No one knows for sure. But as we did for 2017, we asked top scientists and thought leaders in innovation what they expect to see in the new year. Here, lightly edited, are their predictions. Dr. Sean Carroll is a theoretical physicist at the California Institute of Technology in Pasadena.

Azeem Azhar is a strategist, product entrepreneur, and analyst living in London. He is the curator of the weekly newsletter Exponential View, from which the following is adapted. This is the first year I am presenting predictions for the coming year. I've received some incredibly helpful comments from readers via Twitter. This has encouraged me to stick my head above the parapet.

In tracking artificial intelligence trends, Gartner estimates that AI technologies will be in almost every new software product and service by 2020. What's more, 30 percent of CIOs said that AI will be a top five investment priority in 2020. But with any emerging technology there are numerous challenges to watch out for. One of the biggest challenges for enterprises that plan to invest in AI is the staffing element. Artificial intelligence trends tell us that it is an extremely competitive field and the average salaries for machine learning candidates with the right skills and knowledge start at $300,000 per year.

They were the mischievous creatures blamed for causing mechanical failures and faults on aircraft during World War Two - before starring in a hit film as destructive monsters. Now, the gremlins are back - as a new type of killer flying drone. The US Defense Advanced Research Projects Agency (DARPA) research arm is pitting Dynetics and General Atomics (maker of the Predator drone) against each other in a contest to make the craft. Darpa said the program has been deliberately named Gremlins after the imps that British pilots during Wold War Two adopted as their good luck charms. The program envisions launching groups of UASs from existing large aircraft such as bombers or transport aircraft - as well as from fighters and other small, fixed-wing platforms - while those planes are out of range of adversary defenses.

Last year, a strange self-driving car was released onto the quiet roads of Monmouth County, New Jersey. The experimental vehicle, developed by researchers at the chip maker Nvidia, didn't look different from other autonomous cars, but it was unlike anything demonstrated by Google, Tesla, or General Motors, and it showed the rising power of artificial intelligence. The car didn't follow a single instruction provided by an engineer or programmer. Instead, it relied entirely on an algorithm that had taught itself to drive by watching a human do it. Getting a car to drive this way was an impressive feat.

Robots could treble China's bomb and shell production capacity in less than a decade according to a senior scientist involved in a programme that is using artificial intelligence to boost the productivity of ammunition factories. Xu Zhigang, a researcher with the Chinese Academy of Sciences' Shenyang Institute of Automation and a lead scientist with China's "high-level weapon system intelligent manufacturing programme", told the South China Morning Post last Wednesday that about a quarter of the country's ammunition factories had replaced many workers with "smart machines" or begun to do so. The robots, with man-made "hands and eyes", could assemble different types of deadly explosives including artillery shells, bombs and rockets, he said. They could also make more sophisticated ammunition such as guided bombs, equipped with computer chips and sensors, that could carry out precision strikes. They were five times as productive as a human worker, Xu said, but logistical factors such as the supply of raw materials meant the overall productivity boost would fall between 100 to 200 per cent "at a minimum" once all China's ammunition factories were upgraded in the next decade, Xu said.

Central bankers around the world have set up or are creating departments to embrace big data in the quest for deeper insight into the economies they manage. "Isaac Asimov once said, 'I do not fear computers. I fear the lack of them,'" David Hardoon, chief data officer at the Monetary Authority of Singapore, said in a recent speech. "We are now starting to put in place the necessary tools, infrastructure and skillsets to harness the power of data science to unlock insights, sharpen surveillance of risks, enhance regulatory compliance and transform the way we do work." Authorities like Hardoon are tapping publicly-available sources such as Google Trends and jobs websites to help "nowcast" their economies, and confidential data like credit registers that can help identify a stressed bank.

At the World Government Summit in Dubai in February, Tesla and SpaceX chief executive Elon Musk said that people would need to become cyborgs to be relevant in an artificial intelligence age. He said that a "merger of biological intelligence and machine intelligence" would be necessary to ensure we stay economically valuable. Soon afterwards, the serial entrepreneur created Neuralink, with the intention of connecting computers directly to human brains. He wants to do this using "neural lace" technology – implanting tiny electrodes into the brain for direct computing capabilities. Various forms of BCI are already available, from ones that sit on top of your head and measure brain signals to devices that are implanted into your brain tissue.