The common thread across all AI technologies is the ability to impart human-like decision-making capabilities into applications and systems. Artificial intelligence (AI) refers to the simulation of human intelligence in systems programmed to think like humans and mimic their actions. AI includes a broad range of technologies, including cognitive computing, deep learning, expert systems, machine learning, natural language processing, and IBM Watson. The common thread across these areas, and all of AI, for that matter, is the ability to impart human-like decision-making capabilities into applications and systems. Experts predict AI will be rapidly adopted because they believe it will be a disruptive technology across many industries.

There are endless resources for someone who wants to learn to train a deep learning model, but running a successful deep learning project requires managing many additional moving parts that are much less discussed. This talk contributes to filling that gap in our deep learning education resources. Deep learning projects require managing large datasets, heavy-duty dependencies, complex experiments, and large amounts of code. This talk provides best practices for accomplishing these tasks efficiently and reproducibly. Tools that are covered include the Creevey library for processing large collections of files; pip-tools and nvidia-docker for managing dependencies; and MLflow Tracking for tracking experiments.

Supercomputers, like this one at Lawrence Livermore National Laboratory, are designed to tackle the world's toughest computing challenges. Microsoft has built an enormous supercomputer for artificial intelligence work, a new direction for its Azure cloud computing service. The machine has 285,000 processor cores boosted by 10,000 graphics chips for OpenAI, a company that wants to ensure AI technology helps humans. Microsoft announced the machine at its Build conference for developers on Tuesday. Supercomputers, the most powerful computing machines on the planet, are typically used for the most taxing problems. That includes jobs like simulating nuclear weapons explosions, predicting the Earth's future climate and more recently, seeking drugs to fight the coronavirus.

The proliferation of Artificial Intelligence (AI) in the Healthcare sector is one advancement that is worth a watch. Several major companies including big techs are moving forward in the same direction to revolutionize how care is being given to those in need. Recently, a collaboration between Google's DeepMind and Moorfields Eye Hospital NHS Foundation Trust has come up with a development of an AI model that has the potential to predict whether a patient will develop wet AMD within six months. In the future, this system could potentially help doctors plan studies of earlier intervention, as well as contribute more broadly to the clinical understanding of the disease and disease progression. Age-related macular degeneration (AMD) is the biggest cause of sight loss in the UK and the USA and is the third-largest cause of blindness across the globe.

In recent years, we have witnessed the success of modern machine learning (ML) models. Many of them have led to unprecedented breakthroughs in a wide range of applications, such as AlphaGo beating a world champion human player or the introduction of autonomous vehicles. There has been continuous effort, both from industry and academia, to extend such advances to solving real-life problems. However, converting a successful ML model into a real-world product is still a nontrivial task. Firstly, modern ML methods are known for being data-hungry and inefficient.

Deep clustering (DC) has become the state-of-the-art for unsupervised clustering. In principle, DC represents a variety of unsupervised methods that jointly learn the underlying clusters and the latent representation directly from unstructured datasets. However, DC methods are generally poorly applied due to high operational costs, low scalability, and unstable results. In this paper, we first evaluate several popular DC variants in the context of industrial applicability using eight empirical criteria. We then choose to focus on variational deep clustering (VDC) methods, since they mostly meet those criteria except for simplicity, scalability, and stability. To address these three unmet criteria, we introduce four generic algorithmic improvements: initial $\gamma$-training, periodic $\beta$-annealing, mini-batch GMM (Gaussian mixture model) initialization, and inverse min-max transform. We also propose a novel clustering algorithm S3VDC (simple, scalable, and stable VDC) that incorporates all those improvements. Our experiments show that S3VDC outperforms the state-of-the-art on both benchmark tasks and a large unstructured industrial dataset without any ground truth label. In addition, we analytically evaluate the usability and interpretability of S3VDC.

We use convolutional neural networks (CNNs) to analyze authorship questions surrounding the works of Leonardo da Vinci -- in particular, Salvator Mundi, the world's most expensive painting and among the most controversial. Trained on the works of an artist under study and visually comparable works of other artists, our system can identify likely forgeries and shed light on attribution controversies. Leonardo's few extant paintings test the limits of our system and require corroborative techniques of testing and analysis.

Ludwig van Beethoven composed his symphonies between 1799 and 1825, when he was writing his Tenth symphony. As we dispose of a great amount of data belonging to his work, the purpose of this paper is to investigate the possibility of extracting patterns on his compositional model from symbolic data and generate what would have been his last symphony, the Tenth. A neural network model has been built based on the Long Short-Therm Memory (LSTM) neural networks. After training the model, the generated music has been analysed by comparing the input data with the results, and establishing differences between the generated outputs based on the training data used to obtain them. The structure of the outputs strongly depends on the symphonies used to train the network.

Achieving the Bayes optimal binary classification rule subject to group fairness constraints is known to be reducible, in some cases, to learning a group-wise thresholding rule over the Bayes regressor. In this paper, we extend this result by proving that, in a broader setting, the Bayes optimal fair learning rule remains a group-wise thresholding rule over the Bayes regressor but with a (possible) randomization at the thresholds. This provides a stronger justification to the post-processing approach in fair classification, in which (1) a predictor is learned first, after which (2) its output is adjusted to remove bias. We show how the post-processing rule in this two-stage approach can be learned quite efficiently by solving an unconstrained optimization problem. The proposed algorithm can be applied to any black-box machine learning model, such as deep neural networks, random forests and support vector machines. In addition, it can accommodate many fairness criteria that have been previously proposed in the literature, such as equalized odds and statistical parity. We prove that the algorithm is Bayes consistent and motivate it, furthermore, via an impossibility result that quantifies the tradeoff between accuracy and fairness across multiple demographic groups. Finally, we conclude by validating the algorithm on the Adult benchmark dataset.

Few-shot meta-learning has been recently reviving with expectations to mimic humanity's fast adaption to new concepts based on prior knowledge. In this short communication, we give a concise review on recent representative methods in few-shot meta-learning, which are categorized into four branches according to their technical characteristics. We conclude this review with some vital current challenges and future prospects in few-shot meta-learning.