Faculty Summit 2016

Contact: Alex Wade

Microsoft Academic services includes a set of APIs and data that make it easier to build robust apps, and tap into rich, academic data. Plus, you can use the set of productivity tools and services that make it easy to stay-up-to-date on the latest research papers, people, journals, conferences and news.

Contact: Anna Roth

Microsoft Cognitive Services API collection lets you tap into an ever-growing collection of powerful AI algorithms developed by experts in their fields from Microsoft Research—intelligences like vision, speech, language, and knowledge. They simplify a variety of AI-based tasks, giving you a quick way to add top-of-the-line intelligence technologies to your apps with just a few lines of code. Simply drop the REST API call into your app’s code and you’re set.

Contacts: Katja Hofmann and Tim Hutton

We present Project Malmo, an open source AI experimentation platform, designed to support fundamental research in artificial intelligence. Recent research in machine learning, cognitive science, neuro-science and other related areas have led to a renewed interest in general artificial intelligence. With the Project Malmo platform, Microsoft aims to provide an experimentation environment in which promising approaches can be systematically and easily compared, and that fosters collaboration between researchers. The Project Malmo platform is designed to provide a powerful and flexible platform for empirical research that pushes towards addressing fundamental AI research challenges, such as the role of different types of memory, integration of multi-model, high-dimensional sensory data, life-long learning, just to name a few. Project Malmo achieves flexibility by building on top of Minecraft, a popular computer game with millions of players. The game is particularly appealing due to its open ended nature, collaboration with other players, and creativity in game-play. Exactly these same characteristics make it so interesting for AI research. They allow experimenters to create training and test scenarios that range from simple reactive tasks to complex interaction with other agents and even human players. In this demo, we showed the capabilities of the Project Malmo platform, and the kind of research they can enable. These range from 3D navigation tasks to interactive scenarios where agents compete or collaborate to achieve a goal. We gave a brief look under the hood of our system, and showed how new tasks and agents can be easily implemented and experimented with.

Contact: Dave Wecker

LIQUi|> is a software architecture and toolkit for quantum computing. It includes a programming language, optimization and scheduling algorithms, and quantum simulators. The Academic release of LIQUi|> includes state-of-the-art circuit simulation of up to 23 qubits on a single machine. LIQUi|> can be used to translate a quantum algorithm written in the form of a high-level program into the low-level machine instructions for a quantum device. The toolkit includes a compiler, optimizers, translators, various simulators and a host of examples. The complete platform is freely available directly from GitHub (opens in new tab).

Contact: Jim Holbery

The PTFE Laboratory aims to explore the intersection of printed, textile and flexible electronics to define the next computer user interface frontier. Printed electronic capabilities enable the manufacturing of low cost electronic sensors and is a vehicle for new designs that will scale in a cost-effective manner. Flexible electronics offer an opportunity to shift thinking of the human computer interface through new functionality, enabling applications once not possible.

Contacts: Ivan Tashev and David Johnston

The demo showed the advantages provided by the spatial audio in HoloLens device for augmented and virtual reality scenarios, such as gaming, entertainment and virtual presence. While the human vision has certain limited field of view, restricted in addition by the device, humans can hear and locate sound sources coming from 360 degrees (actually 4 pi steradians). We will demonstrate the abilities of the spatial audio to complement the vision and enhance the overall experience for users of a HoloLens device. During the demo the attendee wears a HoloLens device and plays a short interactive game (spatial sound, vision, gesture, voice – 1 minute), watches short movie segment with virtual surround sound on a virtual screen (1 minute), looks/listens around selected places where we have recorded 3D video and audio (1 minute).

Contacts: Jessica Wolk and Dany Rouhana

We want to know more about nearby things, asking questions like «Where can I go to buy that thing?» or «What is going on around me?» We have mined Web browsing logs, Web pages, and Twitter to find local events and discover which stores sell which products. Our store-finder can find nearby stores that sell a given product, which is useful for local search and Cortana reminders. Our local event detector automatically finds events, both planned and unplanned, and presents summaries in the form of event tweets. Our demonstration showed an interactive map that let users browse the local knowledge that we discovered.

Contact: Tricia Mayer

Microsoft is pleased to announce the launch of the new Microsoft Research website: microsoft.com/research (opens in new tab). Through this new experience, we encourage you to explore the research and innovation thriving in research teams all across the company. This web site is part of an evolving approach to keep you and your students informed of opportunities for engagement, and to provide access to Microsoft researchers and their work. Explore the site and help us evolve it even further by providing your perspective and feedback.

Contact: Tom Ball

From its conception in 2011, the Touch Develop project has been about bringing the joy of learning to program modern devices such as smartphones and tablets, as well as laptops and desktops. The Touch Develop team is proud to be part of the BBC micro:bit project. The BBC micro:bit is a small wearable and programmable mbed-based device that visibly features a 5×5 LED display, accelerometer, compass, buttons, I/O pins, Micro USB plug, Bluetooth Low Energy antenna, ARM Cortex-M0 processor, and battery plug. The micro:bit provides a fun introduction to programming and making – switch on, program it to do something fun – wear it, customize it, Just like Arduino, the micro:bit can be connected to and interact with sensors, displays, and other devices. The first wave of micro:bits will land in UK schools this autumn, with every Year 7 student in the UK receiving a micro:bit, for free. As a partner on the micro:bit project, goals are to provide: (1) a browser-based introductory programming experience for students who have never programmed before; (2) an architecture that allows students to dig deeper to uncover the many capabilities of the micro:bit; (3) materials and a platform to support teachers with the micro:bit in their classrooms.

Contact: Jianlong Fu

Automatic plant recognition engine can be a great benefit to both ordinary people and domain experts. By taking advantage of domain knowledge from experts, the-state-of-the-art deep learning techniques and millions of training images, we propose an expert-level image recognition system for plants. The system enables ordinary people to recognize plants around them and further learn comprehensive background knowledge. More importantly, the system can also help domain experts acquire the species distribution in a region by analyzing user-contributed images in some data sharing website of biology. Based on the exploration on building practical systems, we solve the key issues of 1) how to identify the species to be recognized, 2) how to collect training data, and 3) how to train the recognizer. A unified solution and learning techniques are introduced, including fine-grained deep learning network, large-scale knowledge graph mining, and noise-robust model training. Depending on the close collaboration between Institution of Botany, Chinese Academy of Sciences and MSRA, we will leverage and share the training data and knowledge base from domain experts, and integrate the proposed model/algorithm into Microsoft products. This work will generate three main impacts, including: 1) Research impacts: this research topic (e.g., deep learning, fine-grained image classification) is close to the focus of the top conferences in the area of computer vision and artificial intelligence, such as CVPR, ICCV, IJCAI and AAAI. Besides, the training data, domain ontology and knowledge base (including florescence, geographical distribution and morphological descriptions) can also be used as valuable resources for plant recognition research in academic community, for advancing the state-of-the-arts of plant recognition. 2) Business impacts: the capability of expert-level plant recognition can be integrated into many Microsoft intelligent products such as Cortana or XiaoIce to generate more intelligent services, such as image comments and image chat. This technology can also be embedded into a mobile App with client/cloud integration to provide a convenient way for plant recognition by taking photos via widely available cameras. 3) Social impacts: this engine can help ordinal people gain more knowledge about plants when people having high interests on plants but with limited knowledge. Besides, plant recognition engine enables domain experts to acquire the plant distribution in an efficient way. We will demonstrate the plant recognition system on thousands of species across some typical regions in China and United States.

Contacts: Jim Pinkelman, Kenji Takeda, and Vani Mandava

Whether it’s a computer with more memory, a cluster with hundreds of cores, a big data platform, an internet of things solution, or open-source machine learning at scale, you can achieve more using the cloud. Microsoft Azure for Research awards offer large allocations of cloud computing for your research project, and already supports hundreds of researchers worldwide across all domains.

Contact: Ranveer Chandra

Food requirements are expected to double by 2050 to meet the demands of the world population, but the amount of land fit for agriculture is shrinking. Data-driven techniques, such as precision agriculture, could help meet the increased demand. In this demo, we presented FarmBeats, an agricultural IoT system that uses a combination of unmanned aerial vehicles (UAVs) and wireless sensors to enable data-driven agricultural techniques. In doing so, we develop novel algorithms to maximize the coverage of UAV flights given limited battery power, to combine information from a UAV’s video and ground sensor data, and, finally, to achieve cloud connectivity of the farm’s monitoring system while respecting the harsh bandwidth constraints imposed by the farm’s backhaul link to the Internet.

Contacts: Ranvijay Kumar and Sumit Gulwani

The development of this technology was led by Sumit Gulwani at MSR for 5+ years, and then recently (July 2015) CVP Joseph Sirosh further funded development and productization of this technology by setting up a new team inside C&E. This team includes researchers, engineers, UX researchers, and program managers, and is working towards the goal of integrating these technologies inside a dedicated data wrangling product. Sumit moved to this team as part of the MSR rotation program. This technology should be of interest to academics from many different research communities. Building useful end-to-end PBE systems requires cross-disciplinary inspiration. Logical reasoning techniques of the kind developed in the Formal methods community drives development of efficient search algorithms and heuristics. Language design principles from the community of Programming Languages inspires creation of useful DSLs. Machine learning helps in ranking the synthesized programs. And, the field of Human-computer interaction plays a significant role in designing user interaction models. For a recent survey article that cites these various connections, see here (opens in new tab). We have developed various capabilities for data wrangling that benefits from using by-example interaction. Developing a separate PBE technology for each such capability would require significant time, research expertise, and engineering investment. We, however, observed that there are several common ingredients between various PBE technologies, and therefore we refactored them into a core framework (also called PROSE) that has thus far facilitated an order-of-magnitude improvement in creation of the individual PBE technologies and their maintenance. We recently released an academic v0.1 release of the PROSE SDK along with various PBE technologies developed on top of it. The SDK can be downloaded by visiting the Microsoft PROSE website (opens in new tab). We also released our benchmarks to benefit the academia and facilitate further research in the area.

Contact: Daan Liejen

Madoko is a scholarly markdown processor for writing professional articles, books, manuals, and presentations with a focus on simplicity and plain text readability.  The main advantage is that you can write your article in markdown and get both beautiful PDF (through LaTeX) and reflowable HTML: it is time to leave fixed page widths and columns behind and enable better reading of academic content on screens, tablets, and phones. Moreover, Madoko fully supports any LaTeX mathematics and packages using high quality vector graphics in HTML; The online environment at Madoko (opens in new tab) enables seamless cooperation through Dropbox, GitHub and OneDrive.

Contacts: Cem Aykan and Johannes Gehrke

The Office Graph represents a collection of content and activity, and the relationships between them that happen across the entire Office suite. From email, social conversations, meetings and documents, the Office Graph maps the relationships among people and information, and acts as the foundation for rich personalized Office experiences. The Office Graph uses sophisticated machine learning techniques to connect people to the relevant content, conversations and people around them. By tapping into the Office Graph, Delve provides a natural way for users to navigate, discover, and search people and information across an organization.

Contact: Nathan Barnett

Microsoft is leveraging enterprise collaboration data through the Office Graph to achieve double-digit productivity improvements at the world’s largest companies.

Contact: Eduardo Cuervo

Social engagement is a critical element for success in the classroom. Microsoft Embedded Social is an easy way to integrate social features into your learning applications. This Azure hosted service takes the headache out of adding user posts, comments, likes, feeds and other social features into your experience. We take care of all the storage and scalability issues for you. Now you can enable a social experience inside your learning application and engage your classroom.

Contact: Eric Chang

Increasing amounts of data are being collected that can dramatically improve health and wellbeing. In this demo, we highlighted two projects: analyzing data from Microsoft Band and medical image analysis. For Microsoft Band, currently Microsoft Band collects data during sleep but is not able to classify detailed stages such as REM vs. NREM stages. Sleep tracking is the #1 used feature for Microsoft Band and having high quality sleep stage analysis will provide a differentiated feature set for the product. Longer term, we plan to study sleeping habits vs. sleep quality over a large population to develop recommendations for obtaining better sleep quality. In collaboration with the Microsoft Band team, we have already collected sleep clinic data from 40 subjects in Beijing. Currently we are conducting research to create algorithms that can classify REM vs. NREM sleep based on heart beat signals and other signals. We presented our current results and discussed possible future approaches. Team members: Yan Xu, Xudong Ni, Haithem Albadawi, and Eric Chang

For medical image analysis, specifically pathology scan analysis: Pathology images are high in resolution and exhibit patterns in multiple scales. We are analyzing colon cancer pathology scans and utilize big data machine learning approaches to develop techniques for classification and segmentation. These automated techniques will be used to analyze more pathology scans to find correlation between occurrence of cancer cells and final patient outcome. We have achieved state of art results on both colon cancer and brain tumor pathology scans and won the MICCAI 2014 Brain Tumor Digital Pathology Challenge. We have also started working with Intellectual Ventures and Gates Foundation on using similar technology for malaria infection analysis from blood smear samples with very promising results.

Contacts: Jaime Teevan and Shamsi Iqbal

From its conception in 2011, the Touch Develop project has been about bringing the joy of learning to program modern devices such as smartphones and tablets, as well as laptops and desktops. The Touch Develop team is proud to be part of the BBC micro:bit project. The BBC micro:bit is a small wearable and programmable mbed-based device that visibly features a 5×5 LED display, accelerometer, compass, buttons, I/O pins, Micro USB plug, Bluetooth Low Energy antenna, ARM Cortex-M0 processor, and battery plug. The micro:bit provides a fun introduction to programming and making – switch on, program it to do something fun – wear it, customize it, Just like Arduino, the micro:bit can be connected to and interact with sensors, displays, and other devices. The first wave of micro:bits will land in UK schools this autumn, with every Year 7 student in the UK receiving a micro:bit, for free. As a partner on the micro:bit project, goals are to provide: (1) a browser-based introductory programming experience for students who have never programmed before; (2) an architecture that allows students to dig deeper to uncover the many capabilities of the micro:bit; (3) materials and a platform to support teachers with the micro:bit in their classrooms.

Contacts: Olivier Fontana, Tanvi Surti, and Will Lewis

Microsoft Translator’s ecosystem of supported products and apps empowers individuals, organizations and educators to break down language barriers. 

Attendees were able to:

• Communicate with the world through text and speech on any laptop or device in 50+ languages
  • Have a chat with someone who speaks Arabic, Spanish & French over Skype
• Try our translation apps for various platforms
• See how to integrate Speech Translation into your research (or apps)
  • We give you an API (the world’s first Speech Translation API!)
  • We give you sample code (C#, Swift, Java, Python, etc.)
  • Audio input in any of 8 languages, output in 50+
• Watch how Speech Transcription and Translation can change people’s lives!!

Contacts: Ganesh Ananthanarayanan and Victor Bahl

Cities worldwide are deploying millions of cameras for planning and security purposes, and need near real-time video analytics. State-of-the-art video analytics systems are costly (custom-built) and often require manual intervention. Large-scale automated video analytics is a grand challenge for the research community and big data systems. Privacy regulations and bandwidth constraints dictate that video be analyzed across both edge and cloud clusters. Our hybrid edge-cloud video analytics infrastructure: (1) A pilot system in collaboration with Bellevue City which uses traffic cameras for smart city planning of traffic lights, crosswalks, and bike lanes. (2) Resource management for multi-tenant video processing clusters that supports processing pipelines with varying resource-accuracy profiles and delay tolerance. The system runs multiple simultaneous queries and shows outputs of tracking objects (vehicles, pedestrians, bikes), classifying (using a DNN), and counting.

Contact: Sriram Rao

The Cloud and Information Services Lab is an applied research lab working on Systems and Machine Learning Big Data technology and carries out innovative research by building real systems, publishing papers, and contributing to open-source. We enable big data systems in Microsoft products and services and collaborate extensively with MSR, and academia.

Contacts: Lisa Clawson

After 25 years, Microsoft Research continues to push the boundaries of the field of computing. Today’s opportunities are in many ways more impactful and exciting than those of 25 years ago. During all these years, our research has been built on strong relationships and collaborations with academia. As technology advances and our research focus adjusts, we continue to look for impactful ways to engage with you. Engage with us through our competitions, awards, scholarships, fellowships, and more academic resources by visiting us online (opens in new tab).

Contact: Ossie Roycroft and Amanda Young

Interns put inquiry and theory into practice. Alongside fellow doctoral candidates and some of the world’s best researchers, interns learn, collaborate, and network for life. Graduate student interns join us from disciplines spanning computer science, art, sociology and information science, and they contribute to the development, design and study of social computing systems. Interns not only advance their own careers, but they also contribute to the exciting R&D strides our researchers are making.Our TnR Fellowship Program is a two-year fellowship program for PhD students at participating U.S. or Canadian universities enrolled in Computer Science, Electrical Engineering, or Mathematics.

Contacts: Margie Strite and Yesh Naidu

The Microsoft Store device bar featured the latest and greatest PC devices. Everything from the Surface Book to the Dell XPS 15 was available for hands-on demos. The Microsoft Store can help you with all of your technology needs—free workshops, complimentary Answer Desk appointments, and best-in-class customer service.