From Meta-Gradients to Clockwork VAEs, a Global Workspace Theory for Neural Networks and the Edge of Training Stability
Welcome to the April edition of the ‚Machine-Learning-Collage‘ series, where I provide an overview of the different Deep Learning research streams. So what is this series about? Simply put, I draft one-slide visual summaries of one of my favourite recent papers. Every single week. At the end of the month all of the visual collages are collected in a summary blog post. Thereby, I hope to give you a visual and intuitive deep dive into some of the coolest trends. So without further ado: Here are my four favourite papers that I read in March 2021 and why I believe them to be important for the future of Deep Learning. Read More
Monthly Archives: March 2021
Adversarial training reduces safety of neural networks in robots: Research
There’s a growing interest in employing autonomous mobile robots in open work environments such as warehouses, especially with the constraints posed by the global pandemic. And thanks to advances in deep learning algorithms and sensor technology, industrial robots are becoming more versatile and less costly.
But safety and security remain two major concerns in robotics.
… But adversarial training can have a significantly negative impact on the safety of robots, the researchers at IST Austria, MIT, and TU Wien discuss in a paper titled “Adversarial Training is Not Ready for Robot Learning.” Their paper, which has been accepted at the International Conference on Robotics and Automation (ICRA 2021), shows that the field needs new ways to improve adversarial robustness in deep neural networks used in robotics without reducing their accuracy and safety. Read More
The hidden fingerprint inside your photos
They say a picture is worth a thousand words. Actually, there’s a great deal more hidden inside the modern digital image, says researcher Jerone Andrews.
… When you take a photo, your smartphone or digital camera stores “metadata” within the image file. This automatically and parasitically burrows itself into every photo you take. It is data about data, providing identifying information such as when and where an image was captured, and what type of camera was used.
…But metadata is not the only thing hidden in your photos. There is also a unique personal identifier linking every image you capture to the specific camera used. Read More
New AWS tool uses machine learning to watch your services and data for anomalies
AWS has made available Amazon Lookout for Metrics, a service that uses machine learning (ML) to automatically monitor various metrics across business and operational data, detect anomalies and alert the user so they can take appropriate action.
According to AWS, Lookout for Metrics is based on technology used by Amazon itself in business operations, and so reflects 20 years of the firm’s experience in anomaly detection and machine learning. It was built to allow developers to set up autonomous monitoring of important metrics to detect anomalies and identify their root cause in a matter of few clicks. This, AWS claimed, would make it easier to diagnose the root cause of anomalies such as unexpected dips in revenue, high rates of abandoned shopping carts, spikes in payment transaction failures, or increases in new user sign-ups. Read More
YouTuber Creates Roadside AI-Powered Camera To Compliment Dogs That Pass By
Every dog is the best dog. It doesn’t matter whose dog they are, what breed, or how well-behaved they are, they are all good dogs. But do you ever wish that you could tell the dogs how amazing they are, constantly?
Fear not, as YouTuber Ryder Calm Down has the answer. Using a megaphone, a camera, and a smart integrated machine-learning system, the nifty technology and comedy commentator created a device that recognizes dogs as they walk down the street and shouts compliments to them. After all, they deserve it. Read More
This AI Can Generate Convincing Text—and Anyone Can Use It
Some of the most dazzling recent advances in artificial intelligence have come thanks to resources only available at big tech companies, where thousands of powerful computers and terabytes of data can be as copious as free granola bars and nap pods.
A new project aims to show this needn’t be the case, by cobbling together the code, data, and computer power needed to reproduce one of the most epic—and potentially useful—AI algorithms developed in recent years.
Eleuther is an open source effort to match GPT-3, a powerful language algorithm released in 2020 by the company OpenAI that is sometimes capable of writing strikingly coherent articles in English when given a text prompt.
Eleuther is still some way from matching the full capabilities of GPT-3, but last week the researchers released a new version of their model, called GPT-Neo, which is about as powerful as the least sophisticated version of GPT-3. Read More
Fireside Chat with Andrew Ng (2021)
Andrew Ng: Forget about building an AI-first business. Start with a mission.
An AI pioneer reflects on how companies can use machine learning to transform their operations and solve critical problems.
Andrew Ng has worn many hats in his life. You may know him as the founder of the Google Brain team or the former chief scientist at Baidu. You may also know him as your own instructor. He has taught countless students, curious listeners, and business leaders about the principles of machine learning through his wildly popular online courses.
Now in his latest venture, Landing AI, which he started in 2017, he is exploring how businesses without giant data sets to draw on can still join in the AI revolution.
On March 23, Ng joined MIT Technology Review’s virtual EmTech Digital, our annual AI event, to share the lessons he’s learned. Read More
9 Comprehensive Cheat Sheets For Data Science
Sometimes we need a short and to-the-point resource.
In this age of technology, if you ever need to find information about any topic — tech-related or not — you can head to Google, and you will find thousands of materials, articles, books, and videos about that topic. Although this easy access to information had allowed many people worldwide to learn new skills, start a new career and explore topics from the comfort of their home, sometimes the massive amount of information can be overwhelming.
When you look for something and end up with so much information, it can get frustrating and confusing because you don’t know where to start, and at the beginning, it is difficult to see the big picture. Situations like this have lead to the appearance of cheat sheets.
Cheat sheets are an amazing resource for shortcut information about a certain topic. Often, cheat sheets are useful in many ways, but mainly initially, so you can grasp the main concepts and build stones of the topic you’re searching for. In case you want to refresh your memory and go through a straightforward reminder of the topic’s basics. Read More
Building a Naive Bayes Machine Learning Model to Classify Text
A quick start guide to get you up and running with an easy yet highly relevant NLP project in Python
Natural Language Processing (NLP) is an extremely exciting field. It lies at the confluence of computer science, linguistics and artificial intelligence, and is concerned with the interaction between human language and computers. More specifically: its goal is to understand how to program computers to understand and interpret our natural language.
… Nonetheless, building models to classify natural language is relatively straightforward. It’s a cool exercise because it’s relevant. This is a very real application of ML that you could use in your own projects. Read More
Rick's Cafe AI 