Sunday, March 10 to Tuesday, March 12, 2019
This three day conference will include workshops, poster presentations, parallel sessions and plenary talks. In addition to the traditional open exchange of ideas of an academic conference, a key outcome of this conference is the forging of new collaborations across the traditional divide between industry and academia.
Sunday, March 10, 2019
11:00am-12:00pm Registration
12:00pm Workshop Session 1
A Hands-on Introduction to Deep Learning Using PyTorch Yannet Interian, University of San Francisco
An Introduction to Computational Statistics in Python Brian Spiering, University of San Francisco
2:30pm Workshop Session 2
Sequence to Sequence Models with Attention Ananth Sankar, LinkedIn
Augmenting Data Analysis with Data from the US Census James Livsey, UC Census Bureau
4:30pm Check-in and Registration
5:30pm Opening Plenary Deepak Aggarwal, VP of AI at LinkedIn
Title: AI that creates professional opportunities at scale
Monday, March 11, 2019
8:30-9:00am Registration and Coffee & Pastries
9:00-11:00am Concurrent Sessions
Session 1: Network Analysis I RM 527
chaired by Dane Taylor, University of Buffalo
Local graph analytics: beyond characterizing community structure Michael Mahoney, UC Berkeley
Local and higher-order network analysis David Gleich, Purdue University
Inference for network regressions with exchangeable errors Bailey Fosdick, Colorado State University
Multilayer Network Embedding via multi-node2vec and Applications to Network Neuroscience James Wilson, Univ of San Francisco
TBA Lada Aadamic, Facebook
Session 2: High Dimensional Signal Processing and Machine Learning I Rm 529
chaired by Anna Ma, UCSD
Pushing AI to the Edge: Compressive Deep Learning and the
Internet-of-Things Thomas Strohmer, UC Davis
Heuristic Framework for Multi-Scale Testing of the Multi-Manifold Hypothesis Karamatou Yacoubou Djima, Amherst College
Fast Binary Embeddings and Applications Rayan Saab, UC San Diego
Overparameterized Nonlinear Learning: Gradient Descent Takes the Shortest Path? Mahdi Soltanolkotabi, USC
Gradient-based methods for control and learning of dynamical systems Maryam Fazel, U of Washington
TBA; Sivan Aldor
Session 3: Data Science in Marketing I RM 451
chaired by Hui Lin, Netlify
Forecasting for Facebook Infrastructure and Networking Ginger Holt, Facebook
Information Asymmetry and Relevance of Sponsored Listings in Online Marketplaces Kinshuk Jerath, Columbia
Visual Listening In: Extracting Brand Image Portrayed on Social Media Liu Liu, Univ of Colorado
Common Pitfalls in Data Science Projects Ming Li, Amazon
11:00-11:15am Coffee and Refreshments
11:15-11:30am Welcome Remarks Rm 150
11:30-12:30pm Plenary Erin Ledell, Chief Machine Learning Scientist H20.AI
Title: Automatic Machine Learning
12:30-2:00pm Lunch
2:00-4:00pm Concurrent Sessions
Session 1: Technology and Data Science - Accelerants of Sustainable Development Goals I RM 529
chaired by Radhika Shah, Stanford Angels and Entrepreneurs and Jenna Nicholas, Impact Experience
How Computer Vision and Deep Learning Can Help to Achieve SDGs Sabrina Yuan, SecuritAI
Evaluating network interventions Johan Ugander, Stanford
Jumping the Chasm: Challenges in Tracking Data for Business and Philanthropy’s Contributions to the SDGs Heather Grady, Rockefeller Philanthropy Advisors
Engaging Citizens Online Lodewijk Gelauff, Stanford
Session 2: Deep Learning I RM 150
chaired by Yann Dauphin, Google
Deep Model-based Reinforcement Learning of Control Amy Zhang, Facebook
Reinforcement Learning for Robotics Wojciech Zaremba, OpenAI
AI Neuroscience: can we understand the neural networks we train? Jason Yosinski, Uber
Session 1a: Data Science in Marketing RM 527
Intelligent Kit: Making Commerce Automated for Everyone Helen Mou, Shopify
4:15-5:15pm Plenary Margaret Mitchell, Senior Research Scientist, Google RM 150
Title: Bias in the Vision and Language of Artificial Intelligence
5:30-7:30pm Poster Session and Reception 5th Floor
Tuesday, March 12, 2019
9:00-9:30am Registration and Coffee & Pastries
9:30-11:30am Concurrent Sessions
Session 1: High Dimensional Signal Processing and Machine Learning II Rm 527
chaired by Alexander Cloninger, UCSD
The extended Generalized Haar-Walsh Transform (eGHWT) and Its Applications to Data Science Naoki Saito, UC Davis
Large-scale data techniques for Lyme disease Deanna Needell, UCLA
Sparse recovery with dense models: Closing the gap between theory and practice Bernhard Bodmann, U Houston
Deep learning by design Hrushikesh Mhaskar, Claremont Graduate University
Label-aware dimensionality reduction for marker selection Soledad Villar, NYU
Session 2: Design of Experiments RM 529
chaired by Nathaniel Stevens, University of Waterloo
Improving External Validity of A/B Testing using Jackknife Jiannan Lu, UC Berkeley
Designing experiments on networks Vasiliki Koutra, King’s College London
What studying causal inference taught me about experimentation Sonya Berg, Stitch Fix
Targeting, Ramping and Experimentation @ LinkedIn: Our Journey to the Top of the Pyramid Amir Sepehri, LinkedIn
The Analysis of A/B Tests with Comparative Probability Metrics Nathaniel Stevens, University of Waterloo
Session 3: Deep Learning II RM 155
chaired by Yannet Interian, University of San Francisco
Applying Deep Learning to Article Encoding for Fake News Evaluation Mike Tamir, Uber
Leveraging Inheritance and Deep Learning to Detect Genomic Variation Mario Banuelos, California State University, Fresno
Cross Lingual Understanding (XLU): Scaling NLP Across Languages Veselin Stoyanov, Facebook
An Infinitely Customizable Training Loop Sylvain Gugger, fast.ai
Efficient spatio-temporal reasoning for video understanding Yannis Kalantidis, Facebook
11:45-1:15pm Lunch
1:15-1:30pm Coffee and Refreshments
1:30-3:30pm Concurrent Sessions
Session 1: Technology and Data Science - Accelerants of Sustainable Development Goals II RM 529
chaired by Radhika Shah, Stanford Angels and Entrepreneurs and and Jenna Nicholas, Impact Experience
Utilizing Google Earth Engine for Urban Research: Applying Machine Learning Approaches to Map Built-up Land Cover at Scale
Ran Goldblatt, New Light Technologies
Doing Good on Purpose Rather than Evil by Accident with Tech Jim Fruchterman, Tech Matters
4th Industrial Revolution, AI, Culture and Ethics Phil Mui, Salesforce
Ethics of Data & Danish and European perspectives on data for good and initiatives for the SDGs Søren Jørgensen
Session 2: Network Analysis II RM 527
chaired by John Palowitch, Google
Network models for recommender systems Roxana Pamfil, University of Oxford
Nearly-optimal prediction of missing links in networks Aaron Clauset, U Colorado Boulder
Learning Interpretable Clusters for Large-scale Time Series Data Qing Feng, Facebook
Assembly in Populations of Social Networks Abigail Jacobs, UC Berkeley
Flow-Based Local Graph Clustering with Better Seed Set Inclusion Christine Klymko, Lawrence Livermore National Laboratories
Session 3: Data Science in Marketing II RM 451
chaired by Hui Lin, Netlify
Optimal Product Design by Sequential Experiments in High Dimensions Mingyu Joo, UC Riverside
Randomization for Networked Experiments using Random Dot Product Graphs Yichen Qin, U of Cincinnati
Stories that Scale: Google's Daily Information Needs program Mike Brzozowski, Google
3:45-6:00pm Closing Panels and Reception RM 150
Panel I: Transformative Role of Data Science & AI in Advancing SDGs
State of the world today, a glimpse into the future, smarter data analytics, AI & ethics of data
Moderated by Radhika Shah, Stanford Angels and Entrepreneurs
Mehran Sahami , Stanford
Josh Blumenstock, U.C Berkeley
Tom Kalil, Schmidt Futures
Temina Madon, Atlas AI
Panel II: How do we ensure data is harnessed for good and mitigate unintended consequences?
moderated by Jenna Nicholas, Impact Experience
Sam Hamilton, Visa
Waqar Hasan
Keith Coleman, Tesla Foundation
Linda Sheehan, DiCaprio Foundation
Session Descriptions
Deep Learning
From advances in computer vision such as recognizing street signs, 3D objects, to image classification, to advances in sentiment and NLP processing, deep learning (DL) or deep machine learning is rapidly being developed and applied to many areas of scientific and industry application. DL is largely considered a collection of machine learning algorithms largely based on neural networks to model nonlinear and hierarchical features in a data set. The primary family of DL architectures include Convolution Neural Networks, Recurrent Neural networks and Deep Belief networks. This session will bring together academics and industry practitioners to discuss recent advances and uses of DL.
Network Analysis
Network data arises from diverse systems including online social platforms, biological ecosystems, genetic regulation, consumer economies, and transportation infrastructures, and network-data analytics is pervasive throughout industry and academic research efforts. Because of the inherent high-dimensionality and complex correlation structure of networks, network analysis remains an active area of research and development. This field is extremely interdisciplinary with different communities focusing on various aspects including the statistical and mathematical rigor of analyses, the scalability of algorithms for networks that can potentially contain millions of nodes, and the development of domain-specific techniques catering to the uniques demands of diverse applications. Drawing from an interdisciplinary community, this track brings together ten academic and industry researchers to discuss state-of-the-art methods for network science. Talks will cover a variety of topics including general theory, methodology development, algorithm design, and best-practices for diverse applications. The speakers represent a healthy mix of established names and early-career researchers and hail from statistics, biostatistics, computer science, and applied mathematics backgrounds.
High Dimensional Signal Processing and Machine Learning
The challenge of Big Data is not simply that it is big. Many modern datasets are characterized by high-dimensionality, including data from texts, videos, images, and tweets. Dimensionality poses a serious challenge because the amount of data needed to support classical methods of inference grows very rapidly with the number of variables involved - the so-called “curse of dimensionality.” Over the last few decades, many innovative statistical methods have been proposed for high dimensional settings. For example, many binary classification problems involve data that are not linearly separable. However, a mapping to a higher dimensional space can induce separability. This track will bring together leaders in high dimensional analysis to describe novel ideas from signal processing to tackle the challenges of high dimensional data in machine learning.
Design of Experiments
Amidst the present data revolution decisions are increasingly becoming data-driven. However, much of the insight drawn from this influx of data is correlational, when what is often desired is causal inference. Designed experiments provide a framework for the collection and analysis of data that facilitate this type of inference. In the field of data science, experimentation -- colloquially referred to as A/B testing -- is an extremely effective tool for the development and optimization of products and processes. With speakers from both industry and academia, this session will explore modern problems and methodological advances in the design and analysis of experiments with applications to data science.
Technology and Data Science - Accelerants of the Sustainable Development Goals (SDGs)
In this track we will hear from academics from Stanford and Berkeley, tech industry leaders and other thought leaders who are bringing transformative change in advancing the 17 UN Sustainable Development Goals - global goals which are tackling the biggest challenges of our time and aiming to tackle poverty, hunger, access to quality education, healthcare, water, energy for all, sustainable development, gender equality and social justice as well as conserve our lands and forests. For the first time we have a global normative framework that most governments and the UN have adopted and committed to for comprehensive positive change - these ambitious goals recognize the linkages between the various goals and have the spirit of leaving no one behind. Technology, innovation and data science could be powerful tools for advancing and scaling solutions advancing the SDGs. On the other hand, technology can also make things worse if not designed and deployed in a thoughtful manner.
Data Science in Marketing
Big data, the Internet of Things, and social media have all forever changed the art and science of marketing. Today's marketing decision-making happens in real time, every day. Data science give marketers a hands-free way to quickly and effectively respond to data from a customer or potential buyer and tailor fit a product and buying experiences. Buyers on the fence can automatically receive special incentives to buy, and other customers can be directed to related services based on data-driven insight. This track brings together academic and industry researchers to discuss state-of-the-art methods for data science in marketing. A variety of topics include social media marketing, digital product customer service, market research analytics, networked experiment and multichannel customer analytics.