As deep neural networks continue to revolutionize various application domains, there is increasing interest in making these powerful models more understandable and interpretable, and narrowing down the causes of good and bad predictions. We focus on recurrent neural networks, state of the art models in speech recognition and translation. Our approach to increasing interpretability is by combining a long short-term memory (LSTM) model with a hidden Markov model (HMM), a simpler and more transparent model. We add the HMM state probabilities to the output layer of the LSTM, and then train the HMM and LSTM either sequentially or jointly. The LSTM can make use of the information from the HMM, and fill in the gaps when the HMM is not performing well. A small hybrid model usually performs better than a standalone LSTM of the same size, especially on smaller data sets. We test the algorithms on text data and medical time series data, and find that the LSTM and HMM learn complementary information about the features in the text.

Models obtained by decision tree induction techniques excel in being interpretable.However, they can be prone to overfitting, which results in a low predictive performance. Ensemble techniques are able to achieve a higher accuracy. However, this comes at a cost of losing interpretability of the resulting model. This makes ensemble techniques impractical in applications where decision support, instead of decision making, is crucial. To bridge this gap, we present the GENESIM algorithm that transforms an ensemble of decision trees to a single decision tree with an enhanced predictive performance by using a genetic algorithm. We compared GENESIM to prevalent decision tree induction and ensemble techniques using twelve publicly available data sets. The results show that GENESIM achieves a better predictive performance on most of these data sets than decision tree induction techniques and a predictive performance in the same order of magnitude as the ensemble techniques. Moreover, the resulting model of GENESIM has a very low complexity, making it very interpretable, in contrast to ensemble techniques.

SAN FRANCISCO, CA--(Marketwired - Nov 16, 2016) - TechCode, a global network of startup incubators and entrepreneur ecosystems, today announces the first cohort for its new accelerator program named the Global AI Accelerator. The accelerator will help startups build their product, streamline their supply chain, find and hire the right talent, syndicate and close funding rounds, strengthen their IP portfolio and gain access to global distribution channels. The cohort comprises fifty startups from ten countries, including the United States, China, Brazil, France, Germany, India, Israel, Korea, Mexico and Spain, who are integrating artificial intelligence in their technology, such as voice recognition, machine learning, natural language processing and more. Calibrate Education, Inc. creates tools that will allow students to learn in new ways and develop a deeper understanding of the subject matter Eggcyte builds personal cloud devices that let users store, search and privately share photos and videos Jasper is an AI-powered bot which is an evolution from a freelance marketplace into an on-demand recruiting service Sereneti Kitchen uses its robotic chef, Cooki, and food ecosystem, Foodi, to help users cook fresh, modern cuisine Slick is a motorized 3-axis stabilizer for GoPros, allowing action sports lovers to better capture their experiences Smartypans, a 2017 CES Innovation Award honoree, is a pan that computes nutrition, teaches users how to cook and documents recipes via an accompanying mobile app Solidface enables teams to create 3D designs, edit and collaborate in real time through its 3D CAD solution Sutro monitors water quality in pools or spas and allows users to administer chemicals or connect with a pool technician via a mobile app Titanium Falcon is a smart ring that uses motion or gestures to control apps on smart devices Yono is the first in-ear basal body temperature thermometer to collect data comfortably and accurately Calibrate Education, Inc. creates tools that will allow students to learn in new ways and develop a deeper understanding of the subject matter "International expansion is becoming more of a priority for today's startups, especially within the artificial intelligence field as the technology grows rapidly across the globe," said Luke Tang, Head of the TechCode Global AI Accelerator Program. "Our first-of-its-kind accelerator program enables our startups, who are from all over the world, to tap into our global resources to quickly commercialize their technologies and scale up their business. In doing so, we hope to further advance the artificial intelligence industry and help these promising startups find new ways to solve existing problems."

Thanks to the likes of Google, Amazon, and Facebook, the terms artificial intelligence (AI) and machine learning have become much more widespread than ever before. They are often used interchangeably and promise all sorts from smarter home appliances to robots taking our jobs. But while AI and machine learning are very much related, they are not quite the same thing. You can now play a Pictionary-style game called Quick Draw against Google's AI You can now play a Pictionary-style game called Quick Draw against Google's AI You can now play a Pictionary-style game called Quick Draw against Google's AI You can now play a Pictionary-style game called Quick Draw against Google's AI You can now play a Pictionary-style game called Quick Draw against Google's AI AI is a branch of computer science attempting to build machines capable of intelligent behaviour, while Stanford University defines machine learning as "the science of getting computers to act without being explicitly programmed". You need AI researchers to build the smart machines, but you need machine learning experts to make them truly intelligent.

Microsoft's President and Chief Legal Officer Brad Smith was in town yesterday to share advice for how to boost the entrepreneurial scene and tech industry in Milwaukee and beyond. The Appleton native and Columbia University Law School graduate was here to speak "On the Issues" with Mike Gousha at Marquette University Law School, but his day actually got started before that. In the morning Smith met with the local startup scene. Then came the 12:15 forum when he was interviewed by Gousha. Smith later led a lecture at MU with attorneys as the primary audience on intellectual property law and policy.

The sizeable majority of experts surveyed for this report envision major advances in robotics and artificial intelligence in the coming decade. To what degree will AI and robotics be parts of the ordinary landscape of the general population by 2025? Describe which parts of life will change the most as these tools advance and which parts of life will remain relatively unchanged. These are the themes that emerged from their answers to this question. AI and robotics will be integrated into nearly every aspect of most people's daily lives Many respondents see advances in AI and robotics pervading nearly every aspect of daily life by the year 2025--from distant manufacturing processes to the most mundane household activities. Jeff Jarvis, director of the Tow-Knight Center for Entrepreneurial Journalism at the City University of New York, wrote, "Think'Intel Inside'. By 2025, artificial intelligence will be built into the algorithmic architecture of countless functions of business and communication, increasing relevance, reducing noise, increasing efficiency, and reducing risk across everything from finding information to making transactions. If robot cars are not yet driving on their own, robotic and intelligent functions will be taking over more of the work of manufacturing and moving." Vint Cerf, vice president and chief Internet evangelist for Google, responded, "Self-driving cars seem very likely by 2025. Natural language processing will lead to conversational interactions with computer-based systems. Google search is likely to become a dialog rather than a client-server interaction. The Internet of Things will be well under way by this time and interaction with and among a wide range of appliances is predictable. Third party services to manage many of these devices will also be common."

LIDAR sensors are an essential element in future fully autonomous or semi-autonomous self-driving cars. The system operates on the principle of time-of-flight measurement. A very short laser pulse is transmitted, hits an object, is reflected and detected by a sensor. From the time-of-flight of the laser beam it is possible to calculate the distance to the object. Scanning LIDAR systems scan the surroundings of the car horizontally with a laser beam across a certain angular segment and produce a high-resolution 3D map of the environment.

Let me hear your body talk. A new electronic tattoo picks up on subtle noises inside the human body, including the sound of your heart, muscles and gastrointestinal tract. The skin patch could be used in medical monitoring, to detect irregular heartbeats, for example. It could also act as a human-machine interface to use your voice to control a video games. "Our body generates a lot of different sounds," says Howard Liu at the University of Illinois at Urbana-Champaign.

Ben Dickson is a software engineer and the founder of TechTalks. More posts by this contributor: Why it's so hard to create unbiased artificial intelligence How to facilitate the path to brownfield IoT development Why it's so hard to create unbiased artificial intelligence How to facilitate the path to brownfield IoT development Why it's so hard to create unbiased artificial intelligence The transportation industry is associated with high maintenance costs, disasters, accidents, injuries and loss of life. Hundreds of thousands of people across the world are losing their lives to car accidents and road disasters every year. According to the National Safety Council, 38,300 people were killed and 4.4 million injured on U.S. roads alone in 2015. The related costs -- including medical expenses, wage and productivity losses and property damage -- were estimated at $152 billion.

A mobile phone game that tests spatial navigation skills and has been played by 2.4 million people, has become the largest dementia study in history and raised hopes of a breakthrough in diagnosing the disease. Sea Hero Quest, a collaboration between Alzheimer's Research UK, Deutsche Telekom, game designers Glitchers and scientists, has generated the equivalent of 9,400 years of lab-based research since its launch in May. Experts hope to use the data to create the world's first global benchmark for spatial navigation, one of the first abilities affected by dementia, and to develop the game into an early diagnostic test for the disease, which is the leading cause of death in England and Wales. Dr Hugo Spiers, of University College London, who presented the preliminary findings at the Neuroscience 2016 conference in San Diego, said: "This is the only study of its kind, on this scale, to date. Its accuracy greatly exceeds that of all previous research in this area.