Decentralized and Personalized AI, with Privacy by Design

This session was presented at the MozFest19, London.

This interactive session will seek answers to the questions, “Why Decentralised Artificial Intelligence Will Reinvent The Industry?” & “What are the risks, promises and foundational blocks of secure and private AI?” It will introduce them to the new dawn of Federated Learning having the potential to disrupt Cloud Computing, the dominant computing paradigm of today.

The session is divided into two phases:
1. Learning Phase: By story-boarding, participants would understand the concept and benefits as well as use cases of decentralized AI. There would be quizzes throughout the session and I would follow the andragogy- Learn. Apply. Share. Repeat.
2. Practicing Phase- The Ideathon: Attendees will get into groups to come up with potential solutions under which problems can fall. They would come up as a team to build an idea by brainstorming and prototyping algorithms and presenting it to the rest in the end with one-on-one mentoring by myself.

Transcript

1. Decentralized and Personalized AI, with Privacy by Design Charmi Chokshi, Machine Learning Engineer @CharmiChokshi
2. It’s Diwali in India, let’s have some noise in the house!!!
3. Let’s get to know each other?
4. Find out five similarities with person sitting next to you.*
5. What is the most craziest thing you want to do in the next one year?
6. What is the most coolest project (an AI-based) you have ever worked on?
7. What do you wish to get out of this session?
8. Session plan: • Decentralized Data • Federated Computation ◦ Privacy principles • Federated Learning ◦ Privacy technologies • Ideathon…!!!
60. It’s time for the Ideathon!!!
61. Let’s have a look at our smartphones* *or at our lives
62. Can you spot any app to embed Federated Learning into it?
63. Let’s figure out the Model Learning pipeline
64. 64 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
65. 65 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
66. 66 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
67. 67 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
68. 68 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
69. 69 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
70. 70 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
71. 71 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
72. 72 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
73. 73 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
74. 74 Data acquisition Model Deployment Data Cleaning Feature Engineering Model Validation Model Monitoring Model Selection Model Testing Model Training Hyper parameter tuning
75. Brainstorm How could federated learning be applied to the ways you use a computer* each day? * Laptop, phone, tablet, etc.
76. Finalize Your Idea Spend 5 minutes agreeing as a group what your final idea will be.
77. Finalize Your Idea Design - Develop - Deploy
78. Keep building your Ideas!
79. https://federated.withgoogle.com/
80. Sharing Your Ideas What would you say at the Product Launch Party?
81. Sharing Your Ideas Do not forget to invite us!! @mozillafestival @CharmiChokshi
82. References • Federated Optimization: Distributed Optimization Beyond the Datacenter, Jakub Konecny, H. Brendan McMahan, Daniel Ramage, 2015 • Decentralized Collaborative Learning of Personalized Models over Networks Paul Vanhaesebrouck, Aurélien Bellet, Marc Tommasi, 2017 • Towards federated learning at scale: system design, Keith Bonawitz Hubert Eichner and al., 2019 • https://federated.withgoogle.com/ • https://www.youtube.com/watch?v=89BGjQYA0uE
83. Questions? Suggestions? Comments? Charmi Chokshi, Machine Learning Engineer @CharmiChokshi
84. Thank you! Happy Learning :) Charmi Chokshi, ML Engineer, Shipmnts.com LinkedIn, Twitter: @CharmiChokshi Email: charmichokshi@gmail.com