Software Track
BIO
Alexander works on machine learning and statistical data analysis. This includes application areas ranging from document analysis, bioinformatics, computer vision to the analysis of internet data.
In Alexander’s work he has supervised numerous PhD students and researchers and has written over 200 papers, written one book and edited 5 books. His specialties are kernel methods, such as Support Vector Machines and Gaussian Processes, unsupervised information extraction, fast samplers, user models, deep learning, etc. This includes highly scalable models which work on many TB of data and hundreds of millions of users.
ABSTRACT
Machine Learning has seen rapid growth in its range of applications and capabilities. Many of the results are driven by developments in computing, networks, storage, sensors, and the associated availability of data. In particular, different scaling exponents for different hardware resources (Moore’s law for CPUs, Kryder’s law for disks) have led to a sweet spot for certain algorithms, such as deep learning. Moreover, the proliferation of powerful Deep Learning frameworks such as TensorFlow or Apache MxNet have made it easy to design and invent and deploy sophisticated models with ease. In this talk I will discuss capabilities of frameworks, pertinent questions in terms of statistical modeling, and future challenges arising from the ever increasing amount of data feeding machine learning.
BIO
Jia Li is the Head of R&D, Cloud AI and ML. Before joining Google, she was the Head of Research at Snap, leading the research innovation effort of Snap. Prior to that, Jia led the Visual Computing and Learning Group at Yahoo! Labs. Jia received her Ph.D. degree from the Computer Science Department at Stanford University. She is serving as the Program Chair of the ACM Multimedia Conference 2017, the Associate Editor of the Visual Computer: International Journal of Computer Graphics by Springer and the Computer Vision Foundation Industrial Advisory Board Member. Her work has been widely reported in the media including: The Next Web, Ars Technica, ZDNet, GigaOm, Venture Beat, Mirror, Business Insider, New Scientist and MIT Technology Review in recent years. She was listed as one of the ‘Secret Power Players Who Run Snapchat’ by Business Insider in 2016.
ABSTRACT
In a career spanning academia and Silicon Valley, Jia Li has contributed to some of the most influential datasets in the world and participated AI effort from research innovation to real world problem solving via product impact. Now, as Head of R&D for Google Cloud AI, she’s advocating for a democratized approach to ensure that the compute, data, algorithms and talent behind these technologies reach the widest possible audience.
BIO
Stefano Ermon is currently an Assistant Professor in the Department of Computer Science at Stanford University, where he is affiliated with the Artificial Intelligence Laboratory. He completed his PhD in computer science at Cornell in 2015. His research interests include techniques for scalable and accurate inference in graphical models, large-scale combinatorial optimization, and robust decision making under uncertainty, and is motivated by a range of applications, in particular ones in the emerging field of computational sustainability. Stefano’s research has won several awards, including three Best Paper Awards, a World Bank Big Data Innovation Challenge, and was selected by Scientific American as one of the 10 World Changing Ideas in 2016. He is a recipient of the Sony Faculty Innovation Award and NSF CAREER Award.
ABSTRACT
Many of the recent successes of machine learning have been characterized by the availability of large quantities of labeled data. Nonetheless, we observe that humans are often able to learn with very few labeled examples or with only high level instructions for how a task should be performed. In this talk, I will present some new approaches for learning useful models in contexts where labeled training data is scarce or not available at all. Finally, I will discuss ways to use prior knowledge (such as physics laws or simulators) to provide weak forms of supervision, showing how we can learn to solve useful tasks, including object tracking, without any labeled data.
BIO
Manohar Paluri is currently a Research Lead at Facebook AI research and manages the Computer Vision team in the Applied Machine Learning organization. He is passionate about Computer Vision and in the longer term goal of building systems that can perceive the way humans do. In the process of working on various perception efforts throughout his career he spent considerable time looking at Computer Vision problems in Industry and Academia. He worked at renowned places like Google Research, IBM Watson Research Labs, Stanford Research Institute before helping co found Facebook AI Research directed by Dr. Yann Lecun at Facebook. Manohar spent his formative years at IIIT Hyderabad where he finished his undergraduate studies with Honors in Computer Vision and joined Georgia Tech. to pursue his Ph.D. For over a decade he has been working on various problems related to Perception and has made various contributions through his publications at CVPR, ICCV, ECCV, ICLR, KDD, IROS, ACCV etc. He is passionate about building real world systems that are used by billions of people that bring positive value at scale. Some of these systems are running at Facebook and already have tremendous impact in how people communicate using Facebook.
ABSTRACT
Video is becoming ubiquitous on the web. From the capture and creation to consumption a lot of amazing things are happening. As folks working in AI this poses a new challenge and great opportunity for us. If we can make machines understand video the way humans do then we can unlock a long set of applications. But, to be able to get there we need to solve many challenging problems, some of which are obvious ones that the academic community is solving – Datasets, Action Recognition, Multi-Modal understanding, Temporal aggregation, Modeling appearance and motion, Compression etc. I would like to discuss these direction and also talk about more longer term directions on self-serve content understanding, large label embedding and video summarization and so on.