In the opening session of his 2020 introductory course on deep learning, Alexander Amini, a PhD student at the Massachusetts Institute of Technology (MIT), invited a famous guest: former US President Barack Obama. "Deep learning is revolutionizing so many fields, from robotics to medicine and everything in between," said Obama, who joined the class by video conference. After speaking a bit more on the virtues of artificial intelligence, Obama made an important revelation: "In fact, this entire speech and video are not real and were created using deep learning and artificial intelligence." Amini's Obama video was, in fact, a deepfake--an AI-doctored video in which the facial movements of an actor are transferred to that of a target. Since first appearing in 2018, deepfake technology has evolved from hobbyist experimentation to an effective and dangerous tool. Deepfakes have been used against celebrities and politicians and have become a threat to the very fabric of truth.

Convolutional Neural Networks (CNNs) are a subtype of Artificial Neural Networks (ANNs) mostly used for image classification. CNNs follow the biological principle of the replication of a structure capable of identifying patterns to identify these patterns in different locations. It was inspired by the model of cats' visual system proposed by the Nobel prizes winners Hubel and Wiesel at "Receptive fields, binocular interaction and functional architecture in the cat's visual cortex", published in 1962. One of the works that used this inspiration was the Fukushima's Neocognitron, in 1980, although the word Convolution was not used at the time. Therefore, it is not a coincidence that CNNs are very successful in image recognition.

Google's DeepMind just shared AI-generated predictions about the coronavirus that could help researchers stem the global outbreak

Analysis of the effect of training set size on model performance by artificially subsampling patients at random and retraining the model. This analysis was performed using the circEWS alarm system evaluation policy. We observed that model performance decreases drastically when subsampling to less than 5% of the original training set size, and that the model did not show obvious saturation effects as we move to the full size of the data. A linear model baseline (logistic regression; "LogReg"), a tree-ensemble based method (based on lightGBM, "GBM"; used to construct circEWS), an individual decision tree (based on lightGBM, "DecTree"), and a recurrent neural network ("LSTM") were compared. The Tree models received identical input as given to GBM.

Curious how deep learning solutions are affecting your industry? On Wednesday, March 18th, join Upward, Connecticut Center for Advanced Technology (CCAT), and Connecticut Technology Council (CTC) for Hartford's first "teCTalk" focused on artificial intelligence/machine learning. Learn how machines are being taught to interact with the organic world around them and how this smart technology is working to elevate modern business. An esteemed panel of AI/ML experts across industries, including Upward Citizens GalaxE Solutions, VAANGO, and Saya Life, will navigate participants through the complex topic of deep learning. This event is designed to be interactive: pose your questions to the experts and engage with others in the crowd!

This course is intended for developers who are considering a transition to Data Science or for technical Analysts who are willing to do their first steps in advanced Machine Learning. The dynamic of this course contemplates both theoretical content explanation and practical activities. The main goal is to introduce complex Deep Learning algorithms to participants, and help them implement state of the art solutions in different scenarios guided by the instructor.

Microsoft has discharged DeepSpeed, another profound learning optimization library for PyTorch, that is intended to diminish memory use and train models with better parallelism on existing equipment. As per a Microsoft Research blog post reporting the new framework, DeepSpeed improves PyTorch model preparing through a memory enhancement innovation that expands the number of potential parameters a model can be trained with, utilizes the memory nearby to the GPU, and requires just insignificant changes to a current PyTorch application to be helpful. It's the negligible effect on existing PyTorch code that has the best potential effect. As Machine Learning(ML) libraries become entrenched, and more applications become subject to them, there is less space for new frameworks, and progressively incentive to make existing frameworks more performant and versatile. PyTorch is as of now quick with regards to both computational and advancement speed, yet there's consistently opportunity to get better.

Why you should read it: The book was born from a challenge on LinkedIn, (where Andriy is an influencer and has Top Voice distinction for his reach on that platform). His book doesn't need too much of an introduction; it's the Amazon best seller in its category and probably the best condensed collection of knowledge on the topic. Where you can get it: Buy on Amazon. This book is distributed on the "read first, buy later" principle, which means you can freely download the book, read it, and share it with your friends and colleagues, and if you liked the book or found it useful for your work or studies then buy it. Supplement: You can find the companion wiki and the code examples on Github.

Read an article in Quanta Magazine (New theory cracks open the black box of deep learning) about a talk (see 18: Information Theory of Deep Learning, YouTube video) done a month or so ago given by Professor Naftali (Tali) Tishby on his theory that all deep learning convolutional neural networks (CNN) exhibit an "information bottleneck" during deep learning.

The $2.5M SBIR technology investment will help fund a portable, easy-to-use, automated vulnerability scanning solution to effectively and efficiently assess the cyber security posture of embedded devices. In Phase I, ObjectSecurity successfully developed a working proof of concept of the automated technology, supporting extracting the embedded device firmware via embedded systems ports (JTAG, UART etc.) followed by software vulnerability analysis beyond traditional signature-based assessments. The automated technology uses deep learning to intelligently and adaptively choose the best course of action, which ObjectSecurity researched as part of another prior SBIR contract that focused on traditional enterprise networks.