We are excited to announce the general availability (GA) of Databricks Runtime for Machine Learning, as part of the release of Databricks Runtime 5.3 ML. It offers native integration with popular ML/DL frameworks, such as scikit-learn, XGBoost, TensorFlow, PyTorch, Keras, Horovod, etc. In addition to pre-configuring these popular frameworks, DBR ML makes these frameworks easier to use, more reliable, and more performant. Since we introduced Databricks Runtime for Machine Learning in preview in June 2018, we've witnessed exponential adoption in terms of both total workloads and the number of users. Close to 1000 organizations have tried Databricks Runtime ML preview versions over the past ten months.

Building Deep Learning models without callbacks is like driving a car with no functioning brakes -- you have little to no control over the whole process that is very likely to result in a disaster. In this article, you will learn how to monitor and improve your Deep Learning models using Keras callbacks like ModelCheckpoint and EarlyStopping. A callback is a set of functions to be applied at given stages of the training procedure. You can use callbacks to get a view on internal states and statistics of the model during training. You define and use a callback when you want to automate some tasks after every training/epoch that help you have controls over the training process.

In 2018, Amazon announced a chain of convenience stores named Amazon Go, which are partly automated. The stores use several different technologies like computer vision, deep learning algorithms and sensor fusion. The ceilings of the stores are equipped with multiple cameras and the store shelves have weight sensors attached to them. These weight sensors help in the detection of the items taken by a customer and the item is automatically added to the customer's virtual cart. Introduced on January 22nd 2018, Amazon Go has 18 outlets in the United States.

Explainability or interpretability of AI is a huge deal these days, especially due to the rise in the number of enterprises depending on the decisions made by machine learning and deep learning. Naturally, stakeholders want a level of transparency for how the algorithms came up with their recommendations. The so-called "black box" of AI is rapidly being questioned. For this reason, I was encouraged to learn of IBM's recent efforts in this area. The company's research arm just launched a new open-source AI toolkit, "AI Explainability 360," consisting of state-of-the-art algorithms that support the interpretability and explainability of machine learning models.

In a world first, Insilico Medicine, a Hong Kong-based startup developing deep neural networks for drug discovery, has successfully synthesized and pre-clinically validate a drug candidate in just 25 days, making the drug discovery process, including the designing stage, take about 46 days. According to Insilico's research team and its collaborators at the University of Toronto, the method of designing new kinds of molecules by using a deep generative artificial intelligence (AI) model – called generative tensorial reinforcement learning (GENTRL) – not only set a record time compared to traditional methods but also proved to be 15 times faster than a typical pharma corporation's efficient R&D process. It's worth pointing out, especially for readers unfamiliar with the big pharmaceutical industry, that it takes more than a decade and millions of dollars to discover and develop a drug candidate. What's even more depressing about this inefficient industry that keeps passing off the illusion of innovation for real innovation, is that in the last twenty-plus years the success rate for a drug candidate entering Phase I trials have stagnated at under 10%. Meanwhile, in pre-clinical phases the failure rates for new compounds is over 99%.

Deep reinforcement learning(DRL) has been at the center of some of the biggest breakthroughs of artificial intelligence(AI) in the last few years. However, despite all its progress, DRL methods remain incredibly difficult to apply in mainstream solutions given the lack of tooling and libraries. Consequently, DRL remains mostly a research activity that hasn't seen a lot of adoption into real world machine learning solutions. Addressing that problem requires better tools and frameworks. Among the current generation of artificial intelligence(AI) leaders, DeepMind stands alone as the company that has done the most to advance DRL research and development. Recently, the Alphabet subsidiary has been releasing a series of new open source technologies that can help to streamline the adoption of DRL methods.

For those who are new to AI, Machine Learning, and Deep Learning, I recommend taking a look at the following article entitled "An Introduction to AI." I will refer to Machine Learning and Deep Learning as being subsets of AI. Furthermore, this article is non-exhaustive in relation to potential applications of AI to healthcare and Quantum Computing to various sectors of the economy. The reason for the focus on AI in healthcare is in light of recent articles by a few senior medical practitioners in the US expressing concern about the role of AI in healthcare. Some of the concerns expressed, such as the need for improved sharing of data by healthcare participants including hospitals and ensuring the highest quality in the preparation of data, are entirely valid and I take the view that the need for access to data and sharing of data by hospitals may need to become a matter of political and regulatory concern.

We present an image preprocessing technique capable of improving the performance of few-shot classifiers on abstract visual reasoning tasks. Many visual reasoning tasks with abstract features are easy for humans to learn with few examples but very difficult for computer vision approaches with the same number of samples, despite the ability for deep learning models to learn abstract features. Same-different (SD) problems represent a type of visual reasoning task requiring knowledge of pattern repetition within individual images, and modern computer vision approaches have largely faltered on these classification problems, even when provided with vast amounts of training data. We propose a simple method for solving these problems based on the insight that removing peaks from the amplitude spectrum of an image is capable of emphasizing the unique parts of the image. When combined with several classifiers, our method performs well on the SD SVRT tasks with few-shot learning, improving upon the best comparable results on all tasks, with average absolute accuracy increases nearly 40% for some classifiers. In particular, we find that combining Relational Networks with this image preprocessing approach improves their performance from chance-level to over 90% accuracy on several SD tasks.

In fighting against fake news, many fact-checking systems comprised of human-based fact-checking sites (e.g., snopes.com and politifact.com) and automatic detection systems have been developed in recent years. However, online users still keep sharing fake news even when it has been debunked. It means that early fake news detection may be insufficient and we need another complementary approach to mitigate the spread of misinformation. In this paper, we introduce a novel application of text generation for combating fake news. In particular, we (1) leverage online users named \emph{fact-checkers}, who cite fact-checking sites as credible evidences to fact-check information in public discourse; (2) analyze linguistic characteristics of fact-checking tweets; and (3) propose and build a deep learning framework to generate responses with fact-checking intention to increase the fact-checkers' engagement in fact-checking activities. Our analysis reveals that the fact-checkers tend to refute misinformation and use formal language (e.g. few swear words and Internet slangs). Our framework successfully generates relevant responses, and outperforms competing models by achieving up to 30\% improvements. Our qualitative study also confirms that the superiority of our generated responses compared with responses generated from the existing models.

The choice of sentence encoder architecture reflects assumptions about how a sentence's meaning is composed from its constituent words. We examine the contribution of these architectures by holding them randomly initialised and fixed, effectively treating them as as hand-crafted language priors, and evaluating the resulting sentence encoders on downstream language tasks. We find that even when encoders are presented with additional information that can be used to solve tasks, the corresponding priors do not leverage this information, except in an isolated case. We also find that apparently uninformative priors are just as good as seemingly informative priors on almost all tasks, indicating that learning is a necessary component to leverage information provided by architecture choice.