Deep Learning

Private machine learning done right (Ep. 207)

There are many solutions to private machine learning. I am pretty confident when I say that the one we are speaking in this episode is probably one of the most feasible and reliable.I am with Daniel Huynh, CEO of Mithril Security,  a graduate from Ecole Polytechnique with a specialisation in AI and data science. He worked at Microsoft on Privacy Enhancing Technologies under the office of the CTO of Microsoft France. He has written articles on Homomorphic Encryptions with the CKKS explained series (https://blog.openmined.org/ckks-explained-part-1-simple-encoding-and-decoding/). He is now focusing on Confidential Computing at Mithril Security and has written extensive articles on the topic: https://blog.mithrilsecurity.io/. 
In this show we speak about confidential computing, SGX and private machine learning
 
References
Mithril Security: https://www.mithrilsecurity.io/ 
BindAI GitHub: https://github.com/mithril-security/blindai 
Use cases for BlindAI:Deploy Transformers models with confidentiality: https://blog.mithrilsecurity.io/transformers-with-confidentiality/
Confidential medical image analysis with COVID-Net and BlindAI: https://blog.mithrilsecurity.io/confidential-covidnet-with-blindai/ 
Build a privacy-by-design voice assistant with BlindAI: https://blog.mithrilsecurity.io/privacy-voice-ai-with-blindai/ 
Confidential Computing Explained: https://blog.mithrilsecurity.io/confidential-computing-explained-part-1-introduction/ 
Confidential Computing Consortium: https://confidentialcomputing.io/ 
Confidential Computing White Papers: https://confidentialcomputing.io/white-papers-reports/ 
List of Intel processors with Intel SGX:https://www.intel.com/content/www/us/en/support/articles/000028173/processors.html 
https://github.com/ayeks/SGX-hardware 
Azure Confidential Computing VMs with SGX:Azure Docs: https://docs.microsoft.com/en-us/azure/confidential-computing/confidential-computing-enclaves 
How to deploy BlindAI on Azure: https://docs.mithrilsecurity.io/getting-started/cloud-deployment/azure-dcsv3 
Confidential Computing 101: https://www.youtube.com/watch?v=77U12Ss38Zc 
Rust: https://www.rust-lang.org/ 
ONNX: https://github.com/onnx/onnx
Tract, a Rust inference engine for ONNX models: https://github.com/sonos/tract
 

LIDAR, cameras and autonomous vehicles (Ep. 204)

How does an autonomous vehicle see? How does it sense the road? They are equipped of many sensors, of course. Are they all powerful enough? Small enough to hide them and make your car look beautiful? In this episode I speak about LIDAR, high resolution cameras and some machine learning methods adapted to a minimal number of sensors.
 
Our Sponsors
Ready to advance your career in data science? University of Cincinnati Online offers nationally recognized educational programs in business analytics and information systems. Predictive Analytics Today named UC as the No.1 MS Data Science school in the country and is nationally recognized with a proven track record of placing students at high-profile companies such as Google, Amazon and P&G. Discover more about the University of Cincinnati’s 100% online master’s degree programs at online.uc.edu/obais 
 
Amethix works to create and maximize the impact of the world’s leading corporations and startups, so they can create a better future for everyone they serve. We provide solutions in AI/ML, Fintech, Healthcare/RWE, and Predictive maintenance.
 
References
https://patents.google.com/patent/US20220043449A1/en?oq=20220043449
 
 

Is studying AI in academia a waste of time? (Ep. 202)

Companies and other business entities are actively involved in defining data products and applied research every year. Academia has always played a role in creating new methods and solutions/algorithms in the fields of machine learning and artificial intelligence.However, there is doubt about how powerful and effective such research efforts are.Is studying AI in academia a waste of time?
 
Our Sponsors
Explore the Complex World of Regulations. Compliance can be overwhelming. Multiple frameworks. Overlapping requirements. Let Arctic Wolf be your guide.Check it out at https://arcticwolf.com/datascience
 
Amethix works to create and maximize the impact of the world’s leading corporations and startups, so they can create a better future for everyone they serve. We provide solutions in AI/ML, Fintech, Healthcare/RWE, and Predictive maintenance.
 

Zero-Cost Proxies: How to find the best neural network without training (Ep. 201)

Neural networks are becoming massive monsters that are hard to train (without the "regular" 12 last-generation GPUs).Is there a way to skip that?
Let me introduce you to Zero-Cost proxies
 
 
References
https://www.technologyreview.com/2022/08/05/1056814/automation-ai-machine-learning-automl/
https://iclr-blog-track.github.io/2022/03/25/zero-cost-proxies/

Learning from data to create personalized experiences with Matt Swalley from Omneky (Ep. 197)

In this episode I speak with Matt Swalley, Chief Business Officer of Omneky, an AI platform that generates, analyzes and optimizes personalized ad creatives at scale.
We speak about the way AI is used for generating customized recommendation and creating experiences with data aggregation and analytics. And yes! respecting the privacy of individuals.
 
Links
Grow your business with personalized ads https://www.omneky.com/
Data Science at Home Podcast (Live) https://www.twitch.tv/datascienceathome

Embedded Machine Learning: Part 5 - Machine Learning Compiler Optimization (Ep. 186)

This is the last episode of the series "Embedded ML" and I made it for the bravest :) I speak about machine learning compiler optimization to a much greater detail.
Enjoy the episode!
 
Chat with me
Join us on Discord community chat to discuss the show, suggest new episodes and chat with other listeners!
 
Sponsored by Amethix Technologies
Amethix use advanced Artificial Intelligence and Machine Learning to build data platforms and predictive engines in domain like finance, healthcare, pharmaceuticals, logistics, energy. Amethix provide solutions to collect and secure data with higher transparency and disintermediation, and build the statistical models that will support your business.
 
Links
Amethix Embedded Machine Learning
https://tvm.apache.org/
 

Embedded Machine Learning: Part 4 - Machine Learning Compilers (Ep. 185)

In this episode I speak about machine learning compilers, the most important tools to bridge the gap between high level frontends, ML backends and hardware target architectures.
There are several compilers one can choose. Before that, let's get familiar with what a compiler is supposed to do.
Enjoy the episode!
 
Chat with me
Join us on Discord community chat to discuss the show, suggest new episodes and chat with other listeners!
 
Sponsored by Amethix Technologies
Amethix use advanced Artificial Intelligence and Machine Learning to build data platforms and predictive engines in domain like finance, healthcare, pharmaceuticals, logistics, energy. Amethix provide solutions to collect and secure data with higher transparency and disintermediation, and build the statistical models that will support your business.
 
Links
Amethix Embedded Machine Learning
https://tvm.apache.org/
https://github.com/pytorch/glow
https://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html
 
 
 
 

Embedded Machine Learning: Part 3 - Network Quantization (Ep. 184)

In this episode I speak about neural network quantization, a technique that makes networks feasible for embedded systems and small devices.
There are many quantization techniques depending on several factors that are all important to consider during design and implementation.
Enjoy the episode!
 
Chat with me
Join us on Discord community chat to discuss the show, suggest new episodes and chat with other listeners!
 
Sponsored by Amethix Technologies
Amethix use advanced Artificial Intelligence and Machine Learning to build data platforms and predictive engines in domain like finance, healthcare, pharmaceuticals, logistics, energy. Amethix provide solutions to collect and secure data with higher transparency and disintermediation, and build the statistical models that will support your business.
 
Links
Amethix Embedded Machine Learning
Introduction to quantization on Pytorch
 
 
 

Embedded Machine Learning: Part 2 (Ep. 183)

In Part 2 of Embedded Machine Learning, I speak about one important technique to prune a neural network and perform inference on small devices. Such technique helps preserving most of the accuracy with a model orders of magnitude smaller.
Enjoy the show!
 
 
References
The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks
 

Embedded Machine Learning: Part 1 (Ep.182)

This episode is the first of a series about Embedded Machine Learning. I explain the requirements of tiny devices and how it is possible to run machine learning models.
 
Join us on Discord community chat to discuss the show, suggest new episodes and chat with other listeners!
 
Sponsored by Amethix Technologies
Amethix use advanced Artificial Intelligence and Machine Learning to build data platforms and predictive engines in domain like finance, healthcare, pharmaceuticals, logistics, energy. Amethix provide solutions to collect and secure data with higher transparency and disintermediation, and build the statistical models that will support your business.
 
 
References
https://datascienceathome.com/compressing-deep-learning-models-distillation-ep-104/