Deep neural networks are demonstrating excellent performance on several classical vision problems. However, these networks are vulnerable to adversarial examples, minutely modified images that induce arbitrary attacker-chosen output from the network. We propose a mechanism to protect against these adversarial inputs based on a generative model of the data. We introduce a pre-processing step that projects on the range of a generative model using gradient descent before feeding an input into a classifier. We show that this step provides the classifier with robustness against first-order, substitute model, and combined adversarial attacks. Using a min-max formulation, we show that there may exist adversarial examples even in the range of the generator, natural-looking images extremely close to the decision boundary for which the classifier has unjustifiedly high confidence. We show that adversarial training on the generative manifold can be used to make a classifier that is robust to these attacks. Finally, we show how our method can be applied even without a pre-trained generative model using a recent method called the deep image prior. We evaluate our method on MNIST, CelebA and Imagenet and show robustness against the current state of the art attacks.

Apache Spark and TensorFlow are both open-source projects that have made significant impact in the world of enterprise software in recent years. TensorFlow provides a foundational framework for running distributed numerical computations, such as deep learning algorithms, while Spark is a general Hadoop-like, large-scale data processing framework that's also a popular choice for more traditional machine learning algorithms using MLlib. Google Cloud Platform offers managed services for both Apache Spark, called Cloud Dataproc, and TensorFlow, called Cloud ML Engine. Both of these services deliver the power of their respective open-source frameworks in a managed environment, letting you focus on the data science while we worry about the operations. Intuitively, there is some overlap -- Spark provides a framework for big data computations, and the type of datasets that power TensorFlow algorithms tends to be large.

Emerging machine intelligence capabilities and exploding data volumes could enable IT systems to make inferences and predictions, ushering in a new era of cognitive advances. Artificial intelligence (AI) technologies capable of performing tasks normally requiring human intelligence are increasingly being incorporated into enterprise analytics efforts. Yet the bigger story in cognitive computing is machine intelligence, a collection of advances representing machine learning, deep learning, advanced cognitive analytics, robotic process automation (RPA), and bots, to name a few. Spending on various aspects of machine intelligence is projected to reach nearly $31.3 billion in 2019. It is also becoming a priority for CIOs: In Deloitte's 2016 Global CIO Survey, 1,200 IT executives were asked to identify emerging technologies in which they plan to invest significantly in the next two years, and 64 percent intended to invest in cognitive technologies.

This blog post discusses how to turn your images into text describing what is in them so you can later perform analysis on their contents and topics, all right out of a Jupyter Notebook. An example of when this would be useful is if you are given thousands of tweets, and want to know if the image media has any effect on engagement. Lucky for us, instead of writing our own image recognition tool, the engineers at Amazon, Google, and Microsoft completed this task and made their APIs accessible. Here we'll be using Rekognition, Amazon's deep learning-based image and video analysis tool. This blog serves as an example for how to extract information using different Rekognition operations and is not a replacement for reading the documentation.

When it comes to high-performance deep learning on multiple GPUs (and not to mention, multiple machines) I tend to use the mxnet library. Part of the Apache Incubator, mxnet is a flexible, efficient, and scalable library for deep learning (Amazon even uses it in their own in-house deep learning). Inside the ImageNet Bundle of my book, Deep Learning for Computer Vision with Python, we use the mxnet library to reproduce the results of state-of-the-art publications and train deep neural networks on the massive ImageNet dataset, the de facto image classification benchmark (which consists of 1.2 million images). As scalable as mxnet is, unfortunately it misses some of the convenience functions we may find in Keras, TensorFlow/TensorBoard, and other deep learning libraries. One of these convenience methods mxnet misses is plotting accuracy and loss over time.

We continued to cover interesting and important stories in the intersecting fields of AI, Analytics, Big Data, Data Science, Deep Learning, and Machine Learning. We published over 1200 stories, had over 4.8 million visitors who saw over 14 million pages. Our subscribers have grown to over 200,000 subscribers/followers via email and social media. Here are the top KDnuggets stories from 2017 - read them, if you have not already. How to understand Gradient Descent algorithm, by Jahnavi Mahanta 17 More Must-Know Data Science Interview Questions and Answers, by Gregory Piatetsky 6 Reasons Why Python Is Suddenly Super Popular, by Kayla Matthews 7 More Steps to Mastering Machine Learning With Python, by Matthew Mayo Here are the most shared stories in 2017.

About this course: This course will teach you the "magic" of getting deep learning to work well. Rather than the deep learning process being a black box, you will understand what drives performance, and be able to more systematically get good results. You will also learn TensorFlow. This is the second course of the Deep Learning Specialization.

Deep learning algorithms have benefited significantly from the recent performance gains of GPUs. However, it has been uncertain whether GPUs can speed up powerful classical machine learning algorithms such as generalized linear modeling, random forests, gradient boosting machines, clustering, and singular value decomposition. Today I'd love to share another interesting presentation from #H2OWorld focused on H2O4GPU. H2O4GPU is a GPU-optimized machine learning library with a Python scikit-learn API tailored for enterprise AI. The library includes all the CPU algorithms from scikit-learn and also has selected algorithms that benefit greatly from GPU acceleration. In the video below, Jon McKinney, Director of Research at H2O.ai, discussed the GPU-optimized machine learning algorithms in H2O4GPU and showed their speed in a suite of benchmarks against scikit-learn run on CPUs.

Cray Accel AI Lab Cray is launching a deep learning innovation center – the Cray Accel AI Lab – to serve as the focal point for the ... AI NEWS Global supercomputer leader Cray Inc. (Nasdaq:CRAY) today announced a comprehensive set of Artificial Intelligence (AI) products and ... The evolution of the computer now permits serious consideration of systems that replace, or more ... Every Moment of Your Life – Generational AI Visualization systems are those systems that convey and process information graphically. Here are four AI insights on where I see AI today. Even with the explosive pace of technology, human endeavors in artificial intelligence are ...

I am excited to share a new deep learning model performance trajectory graph. Obviously is going to take some training and practice to read these graphs quickly: but that is petty much true for all visualizations. The methods work with just about any staged machine learning algorithm (neural nets, deep learning, boosting, random forests, and more) and can also be adapted to non-staged bug regularized methods (lasso, elastic net, and so on). The graph is now part of the development version of WVPlots. And we have complete worked examples for Keras and xgboost.