Filtering by: Morning2

End to End Machine Learning with MLDB.ai — Jeremy Barnes (MLDB.ai)
Oct
12
12:00 PM12:00

End to End Machine Learning with MLDB.ai — Jeremy Barnes (MLDB.ai)

MLDB is an open-source database designed for machine learning. You can install it wherever you want and send it commands over a RESTful API to store data, explore it using SQL, then train machine learning models and expose them as APIs. In this talk, we will cover how to build a Predictive API "end to end" from data exploration to model evaluation to deployment, all using only simple MLDB API calls from MLDB's Notebook interface.

Jeremy Barnes is an entrepreneur and technology leader, active at the intersection of artificial intelligence and industry. He has 15 of years of experience applying machine learning to develop innovative products. Prior to MLDB.ai, he co-founded Datacratic, an enterprise software company developing machine learning technology for the marketing industry. Before that, Jeremy co-founded Idilia, a computational linguistics company, where he was responsible for research and development of Idilia's machine learning based core computational linguistics technology.

Linkedin

 

View Event →
BayesDB and VizGPM.js: open-source AI for visually exploring complex databases — Richard Tibbetts (MIT)
Oct
11
12:00 PM12:00

BayesDB and VizGPM.js: open-source AI for visually exploring complex databases — Richard Tibbetts (MIT)

Artificially intelligent data products don’t have to be limited to answering simple natural language queries. Navigation, search, and retrieval of structured data, even by sophisticated domain experts, benefit from using AI to infer data’s latent structure. Using open source BayesDB and VizGPM.js, we demonstrate interfaces for browsing US census and software performance data.

Richard Tibbetts is a software entrepreneur, database and programming languages nerd, a Visiting Scientist at MIT Probabilistic Computing and a leader of the "BayesDB": probcomp.csail.mit.edu/bayesdb open source project. Prior to MIT Richard was founder and CTO at StreamBase, a CEP company that merged with TIBCO in 2013. Richard is also the CEO of Empirical Systems a stealth mode startup.

Twitter: @tibbetts - Linkedin - Website

View Event →
Development and cloud deployment of machine learning models for heartbeat classification on data from wearable devices — Ikaro Silva (MC10)
Oct
11
11:30 AM11:30

Development and cloud deployment of machine learning models for heartbeat classification on data from wearable devices — Ikaro Silva (MC10)

Electrical heart signals are one of the most recorded and stored physiological data in healthcare. With cardiovascular diseases being the single most common cause of death in the world, automatic analysis of cardiac signals under normal ambulatory conditions is expected to play a crucial role in assisting clinicians identify health issues. A critical step towards this goal is the automatic classification of heartbeats. The purpose of this work is to showcase the development and deployment of a cloud system for heartbeat classification collected from wearable devices.

Dr. Ikaro Silva is a Data Scientist at MC10 Inc and is responsible for developing algorithms that process the biological signals collected through MC10's unique wearable form factors. Dr. Silva is also a research scientist at MIT, where he is involved in augmenting PhysioNet's open source software and research.

View Event →
Prediction | Production: Lessons from ‘Over-the-wall’ — Stuart Bailey (Open Data Group)
Oct
11
11:00 AM11:00

Prediction | Production: Lessons from ‘Over-the-wall’ — Stuart Bailey (Open Data Group)

Data Science and Dev Ops teams live on opposite sides of a wall in most organizations. Despite the separation, these teams should work together to develop a coherent process to release analytic products, support those products and maintain sanity. We propose an institutional capability, ‘Analytic Operations’, to support data-driven processes within lines-of-business. We hope to share lessons learned practicing Analytic Ops and present a set of best practices for Analytic Ops teams. We also demo open source tools that reduce frictions between Data Science and Ops/Deployment teams.

Stuart Bailey is a partner and the Chief Technology Officer at the Open Data Group. He is a technologist and entrepreneur who has been focused on analytic and data intensive distributed systems for over two decades. Prior to Open Data Group, Stuart was the founder and most recently Chief Scientist of Infoblox (NYSE:BLOX), a Sequoia Capital-backed company. More than half the Fortune 500 rely on the Infoblox automated distributed system solutions for essential, software-based network control.

Twitter: @stu_bailey - Linkedin

 

View Event →
[Tutorial] Evaluating Failure Prediction Models for Predictive Maintenance — Shaheen Gauher (Microsoft)
Oct
10
11:40 AM11:40

[Tutorial] Evaluating Failure Prediction Models for Predictive Maintenance — Shaheen Gauher (Microsoft)

Predictive Maintenance is about anticipating failures and taking preemptive actions. In the realm of predictive maintenance, the event of interest is an equipment failure. Modelling for Predictive Maintenance falls under the classic problem of modelling with imbalanced data when only a fraction of the data constitutes failure. This kind of data poses several issues. In this talk I will highlight some of the pitfalls and challenges of building a model with such data and describe ways to circumvent the problems using real use cases and examples.

Shaheen Gauher, PhD, is a Data Scientist in Information Management and Machine Learning at Microsoft. She develops end to end data driven advanced analytic solutions for external customers working across all verticals.

Twitter: @Shaheen_Gauher - Linkedin

View Event →
[Tutorial] Beyond Churn Prediction: An Introduction to Uplift Modelling — Pierre Gutierrez (Dataiku)
Oct
10
11:00 AM11:00

[Tutorial] Beyond Churn Prediction: An Introduction to Uplift Modelling — Pierre Gutierrez (Dataiku)

In several industries (e-business, telcos…), a common approach to diminishing user churn is to use machine learning to score individual customers by churn probability and target them with specific messaging or offers. However, this approach may be ineffective since it does not optimize what is called "true lift" or "uplift": the effect of an action on churning probability. This talk aims at introducing uplift modeling in a tutorial like format. We’ll cover the basics of the theory as well as how to make it work in practice. We will illustrate the talk with examples from real life.

Pierre Gutierrez is a senior data scientist at Dataiku. As a data science expert and consultant, Pierre has worked in diverse sectors such as e-business, retail, insurance or telcos. He has experience in various topics such as fraud detection, bot detection, recommender systems, or churn prediction.

Twitter: @prrgutierrez - Linkedin

 

View Event →
[Tutorial] Predicting Customer Behavior — Vinny Senguttuvan (Metis)
Oct
10
11:00 AM11:00

[Tutorial] Predicting Customer Behavior — Vinny Senguttuvan (Metis)

We will look at ways of applying data science and machine learning to better understand customers and improve their user experience. From a practical industry-application perspective, we will discuss the following: measuring popularity, statistical significance in A/B testing, survival analysis, predictive lifetime value and recommendation systems. We will review the concepts and some of the math behind these, while also addressing the real world challenges faced by many of these implementations.

Vinny is a Senior Data Scientist and Professor at Metis. Previously, he was a Lead Data Scientist at High 5 Games and an R&D Programmer at Blue Sky Studios (the animation company that made Ice Age, Rio and Peanuts).

View Event →