Art has long been considered the exclusive domain of human creativity. But turns out machines can do a lot more in the creative realm than we humans can imagine. In October 2018, Christie's sold first AI-generated painting for $432,500. Titled Edmond de Belamy, the artwork was expected to sell for $10,000. Obvious art created this masterpiece using Generative Adversarial Network (GAN) algorithm by feeding the system with 15,000 portraits created between the 14th and 20th century.

How can you tell if an AI technology that's actually part of the AI revolution? What went wrong with artificial intelligence? This transformative technology was supposed to change everything. I've seen first-hand the incredible potential it has--both as a professor of computer science at the University of Michigan and as the founder of Clinc, ZeroShotBot, Myca.ai, a non-profit called ImpactfulAI, and several other AI-focused companies. So, why has it devolved into overhyped solutions, marketing noise, and an endless spin of the same, tired ideas?

It can be difficult to understand exactly what's going on inside of a deep learning model, which is a real problem for companies concerned about bias, ethics, and explainability. Now IBM is developing something called AI FactSheets, which it describes as a nutrition label for deep learning that explains how models work and that can also detect bias. AI FactSheets is a new addition to Watson Open Scale that will provide a plain-language description of what's going on inside deep learning models. The software, which is expected to be generally available soon, can work with AI models developed by Watson Studio, or any other AI model accessible from a REST API. After being exposed to the model, AI FactSheets generates a PDF with information about bias, trust, and transparency aspects of a given deep learning model.

For your capstone project, you'll tackle a real-world data problem from end to end. Develop a pitch and problem statement, source and collect relevant data, conduct an exploratory data analysis, and build a predictive model. You'll document and share your findings through a presentation, technical report, and non-technical summary. Throughout this Accelerator Course you'll also compile a portfolio of projects designed to reinforce what you've learned in each unit. Not only that, You will get to Work on Projects from Topics such as Data Analytics, Data Visualization, Machine Learning, Time-series Forecasting, Deep Learning, Recommendation Engines, NLP, Chatbots, Web Scraping, Computer Vision, etc.

Over the last few years, Voice Assistants have become ubiquitous with the popularity of Google Home, Amazon Echo, Siri, Cortana, and others. These are the most well-known examples of Automatic Speech Recognition (ASR). This class of applications starts with a clip of spoken audio in some language and extracts the words that were spoken, as text. For this reason, they are also known as Speech-to-Text algorithms. Of course, applications like Siri and the others mentioned above, go further.

The Daily Star's FREE newsletter is spectacular! In a move that will terrify anyone who makes music for a living, an Artificial Intelligence researcher has used deep fake technology to recreate a perfect copy of Eminem's 1999 smash "My Name Is…". The AI has duplicated the exact tone and intonation of the Missouri-born rapper but updated his seminal rap to take sideswipes at Billie Eilish, former president Donald Trump and instant dating site Tinder. YouTuber, 30 Hertz, who created the track, says he uses AI to generate "synthetic parody songs and other poorly written material" in a disclaimer that echoes scatalogical 90s comedy show South Park. He's not the only AI tinkerer using artificial intelligence to recreate the works of famous musicians. OpenAI's Jukebox project uses the same technology to create entire songs that might have been recorded by Frank Sinatra, had the Rat Pack legend not died in 1998.

Deep learning has shown a great potential for speech separation, especially for speech and non-speech separation. However, it encounters permutation problem for multi-speaker separation where both target and interference are speech. Permutation Invariant training (PIT) was proposed to solve this problem by permuting the order of the multiple speakers. Another way is to use an anchor speech, a short speech of the target speaker, to model the speaker identity. In this paper, we propose a simple strategy to train a long short-term memory (LSTM) model to solve the permutation problem in speaker separation. Specifically, we insert a short speech of target speaker at the beginning of a mixture as guide information. So, the first appearing speaker is defined as the target. Due to the powerful capability on sequence modeling, LSTM can use its memory cells to track and separate target speech from interfering speech. Experimental results show that the proposed training strategy is effective for speaker separation.

To shift the computational burden from real-time to offline in delay-critical power systems applications, recent works entertain the idea of using a deep neural network (DNN) to predict the solutions of the AC optimal power flow (AC-OPF) once presented load demands. As network topologies may change, training this DNN in a sample-efficient manner becomes a necessity. To improve data efficiency, this work utilizes the fact OPF data are not simple training labels, but constitute the solutions of a parametric optimization problem. We thus advocate training a sensitivity-informed DNN (SI-DNN) to match not only the OPF optimizers, but also their partial derivatives with respect to the OPF parameters (loads). It is shown that the required Jacobian matrices do exist under mild conditions, and can be readily computed from the related primal/dual solutions. The proposed SI-DNN is compatible with a broad range of OPF solvers, including a non-convex quadratically constrained quadratic program (QCQP), its semidefinite program (SDP) relaxation, and MATPOWER; while SI-DNN can be seamlessly integrated in other learning-to-OPF schemes. Numerical tests on three benchmark power systems corroborate the advanced generalization and constraint satisfaction capabilities for the OPF solutions predicted by an SI-DNN over a conventionally trained DNN, especially in low-data setups.

Stochastic gradient descent (SGD) is the main approach for training deep networks: it moves towards the optimum of the cost function by iteratively updating the parameters of a model in the direction of the gradient of the loss evaluated on a minibatch. Several variants of SGD have been proposed to make adaptive step sizes for each parameter (adaptive gradient) and take into account the previous updates (momentum). Among several alternative of SGD the most popular are AdaGrad, AdaDelta, RMSProp and Adam which scale coordinates of the gradient by square roots of some form of averaging of the squared coordinates in the past gradients and automatically adjust the learning rate on a parameter basis. In this work, we compare Adam based variants based on the difference between the present and the past gradients, the step size is adjusted for each parameter. We run several tests benchmarking proposed methods using medical image data. The experiments are performed using ResNet50 architecture neural network. Moreover, we have tested ensemble of networks and the fusion with ResNet50 trained with stochastic gradient descent. To combine the set of ResNet50 the simple sum rule has been applied. Proposed ensemble obtains very high performance, it obtains accuracy comparable or better than actual state of the art. To improve reproducibility and research efficiency the MATLAB source code used for this research is available at GitHub: https://github.com/LorisNanni.

Deep Convolutional Neural Networks (CNNs) have long been the architecture of choice for computer vision tasks. Recently, Transformer-based architectures like Vision Transformer (ViT) have matched or even surpassed ResNets for image classification. However, details of the Transformer architecture -- such as the use of non-overlapping patches -- lead one to wonder whether these networks are as robust. In this paper, we perform an extensive study of a variety of different measures of robustness of ViT models and compare the findings to ResNet baselines. We investigate robustness to input perturbations as well as robustness to model perturbations. We find that when pre-trained with a sufficient amount of data, ViT models are at least as robust as the ResNet counterparts on a broad range of perturbations. We also find that Transformers are robust to the removal of almost any single layer, and that while activations from later layers are highly correlated with each other, they nevertheless play an important role in classification.