Play all

[] Musical introduction to Blaise Thomson

[] From Intractable to Interactable: Unleashing Sensitive Datasets with Distributed Data Science

[] Outline

[] Introduction to Distributed Data Science

[] The context

[] What is going wrong?

[] Distributed Data Science

[] Sending Algorithms to Data

[] How Distributed Data Science solves the problem

[] Usage-based access control

[] Privacy protections

[] Dataset-level privacy-enhancing technologies

[] Collaboration-level privacy-enhancing technologies

[] Confidential Collaboration: Secure Aggregation and Federated Learning

[] Secure Aggregation

[] Federated Learning

[] An example experimental setup Retinal OCT Kaggle

[] An example: Error rates

[] Confidential Collaboration: Private Set Intersection

[] Private Set Intersection - Simple hashing

[] Private Set Intersection - more complex hashing

[] Other use cases

[] Configurations

[] Internal use within the company or group

[] Embedded: Single lead, leveraging partners' data

[] Data Resellers

[] Alliance - Consortium of data providers and data scientists

[] Summary

[] Sign up for the Bitfount Open Beta Launch at https://www.bitfount.com/!

[] Data Errors Obfuscation

[] Histograms with SQL queries on structured or unstructured data

[] Implicit bias detection or model failures

[] Theory vs Application

[] Skipping details approach

[] OpenMined PySyft

[] MLOops!

[] Siri products MLOops!

[] Wrap up

Description:

Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only! Grab it Explore the world of distributed data science and its potential to unlock sensitive datasets in this 39-minute conference talk from the MLOps Community Meetup. Discover how sending algorithms to data can extract value while maintaining strict privacy and security guarantees. Learn about federated machine learning, secure aggregation, and private set intersection techniques. Gain insights into various configurations for implementing distributed data science, from internal company use to data reseller models. Understand the challenges of working with sensitive data and how distributed approaches can improve model quality. Dive into real-world examples, including a retinal OCT experiment, and explore the balance between theory and application in this emerging field. Get introduced to tools like Bitfount's platform and OpenMined PySyft, and hear about potential pitfalls and "MLOops" moments in distributed data science projects.

Unleashing Sensitive Datasets with Distributed Data Science

MLOps.community

Add to list

#Computer Science #Machine Learning #Federated Learning #MLOps #Information Security (InfoSec) #Cybersecurity #Data Privacy #Cryptography #Private Set Intersection