In recent years, with the popularization of deep learning frameworks and large datasets, researchers have started parallelizing their models in order to train faster. This is crucially important, because they typically explore many hyperparameters in order to find the best ones for their applications. This process is time consuming and, consequently, speeding up training improves productivity. One approach to parallelize deep learning models followed by many researchers is based on weak scaling. The minibatches increase in size as new GPUs are added to the system. In addition, new learning rates schedules have been proposed to fix optimization issues that occur with large minibatch sizes. In this paper, however, we show that the recommendations provided by recent work do not apply to models that lack large datasets. In fact, we argument in favor of using strong scaling for achieving reliable performance in such cases. We evaluated our approach with up to 32 GPUs and show that weak scaling not only does not have the same accuracy as the sequential model, it also fails to converge most of time. Meanwhile, strong scaling has good scalability while having exactly the same accuracy of a sequential implementation.

In this project, we developed a deep learning system applied to human retina images for medical diagnostic decision support. The retina images were provided by EyePACS (Eyepacs, LLC). These images were used in the framework of a Kaggle contest (Kaggle INC, 2017), whose purpose to identify diabetic retinopathy signs through an automatic detection system. Using as inspiration one of the solutions proposed in the contest, we implemented a model that successfully detects diabetic retinopathy from retina images. After a carefully designed preprocessing, the images were used as input to a deep convolutional neural network (CNN). The CNN performed a feature extraction process followed by a classification stage, which allowed the system to differentiate between healthy and ill patients using five categories. Our model was able to identify diabetic retinopathy in the patients with an agreement rate of 76.73% with respect to the medical expert's labels for the test data.

Pieter Abbeel UC Berkeley In recent years, deep generative models have been shown to'imagine' convincing high-dimensional observations such as images, audio, and even video, learning directly from raw data. In this work, we ask how to imagine goal-directed visual plans - a plausible sequence of observations that transition a dynamical system from its current configuration to a desired goal state, which can later be used as a reference trajectory for control. We focus on systems with high-dimensional observations, such as images, and propose an approach that naturally combines representation learning and planning. Our framework learns a generative model of sequential observations, where the generative process is induced by a transition in a low-dimensional planning model, and an additional noise. By maximizing the mutual information between the generated observations and the transition in the planning model, we obtain a low-dimensional representation that best explains the causal nature of the data. We structure the planning model to be compatible with efficient planning algorithms, and we propose several such models based on either discrete or continuous states. Finally, to generate a visual plan, we project the current and goal observations onto their respective states in the planning model, plan a trajectory, and then use the generative model to transform the trajectory to a sequence of observations. We demonstrate our method on imagining plausible visual plans of rope manipulation.

Several machine learning frameworks have emerged to streamline the development and deployment of AI applications. These frameworks help abstract away the grunt work of testing and configuring AI workloads for experimentation, optimization and production. However, developers need to make some hard choices when it comes to picking the right framework. Some may want to focus on ease of use when training a new AI algorithm, while others may prioritize parameter optimization and production deployment. Different frameworks have different strengths and weaknesses in these diverse areas.

This video is part of the Udacity course "Deep Learning".

Dubai: In a bid to curb the rising incidence of diabetes, the Dubai Diabetes Centre (DDC) plans to introduce Artificial Intelligence to detect retinopathy, start tele-monitoring of patients who miss their appointments and also introduce obesity clinics in the emirate. Elaborating on the use of Artificial Intelligence Dr M Hamed Farooqi, director of the multidisciplinary centre said: "As per international diabetes standards, we need to have 14 retinal images per diabetic. The estimated number of diagnosed diabetics in the UAE exceeds one million. To interpret 14 million images per year, we need more than 50 eye specialists working full-time. Deep learning system (DLS) using Artificial Intelligence are capable of identifying diabetic retinopathy and related eye diseases using retinal images with a high degree of accuracy. Thus using AI cannot only help provide retinopathy screening for a large number of diabetics but also lead to better utilisation of resources and time of ophthalmologists."

In the future, AI will affect the nature and distribution of jobs and these non-profit bodies bring together a cross-section of industry leaders such as robotics, computing. Even though there has been recent AI wins in terms of self-driving cars, DeepMind's Go victory, we still need to charge ahead for breakthroughs in a slew of public sector areas. An area of interest has been integrative AI approach -- how to combine capabilities in image recognition, speech recognition, computer vision and natural language to explore tougher problems. For example, we are seeing a number of startups that use emerging technologies to tackle social problems in India. Case in point, Fasal an IoT startup gives ML-powered insights to farmers, giving rise to precision agriculture.

Commercially-applied AI has expanded in recent years, driven by a combination of computing power, the availability of huge datasets and advances in machine learning (which includes deep learning). While often used for predictive analytics, as well as image and speech classification, machine learning can be combined with elements of'traditional' AI such as natural language processing, strategic planning and logical reasoning to deliver powerful autonomous agents. So how prevalent is AI? Outside of large tech companies that have been utilising AI in service delivery for a number of years, much of the innovation is still in its infancy and is largely confined to the lab in the form of proof of concepts or R&D. The focus for business now has to be on creating an environment which fosters successful transition into real world value delivery.

Tech leaders, including Elon Musk and the three co-founders of Google's AI subsidiary DeepMind, have signed a pledge promising to not develop "lethal autonomous weapons." It's the latest move from an unofficial and global coalition of researchers and executives that's opposed to the propagation of such technology. The pledge warns that weapon systems that use AI to "[select] and [engage] targets without human intervention" pose moral and pragmatic threats. Morally, the signatories argue, the decision to take a human life "should never be delegated to a machine." On the pragmatic front, they say that the spread of such weaponry would be "dangerously destabilizing for every country and individual."

It is debatable whether Ned Ludd ever existed. But the movement he gave his name to certainly did. In the early 19th century, Luddite mobs of English weavers took to smashing the new automated looms they blamed for threatening their livelihoods. Today, Luddite fears are seeing a resurgence. Admittedly, no one is yet torching factories full of industrial robots or taking their baseball bats to banks of computers running "artificial intelligence" deep learning algorithms.