Deep learning is becoming an increasingly important part of the artificial intelligence (AI) toolkit, yet it is often misunderstood. Although it supports a developing market and is often touted as an important direction for corporate innovation, it needs to be viewed in context. This is a developing area of technology that is rapidly creating its own domain, becoming richer and increasingly varied. We will soon see new opportunities based specifically on deep learning. The market for deep learning solutions continues to expand.

Unsupervised learning, in particular learning general nonlinear representations, is one of the deepest problems in machine learning. Estimating latent quantities in a generative model provides a principled framework, and has been successfully used in the linear case, e.g. with independent component analysis (ICA) and sparse coding. However, extending ICA to the nonlinear case has proven to be extremely difficult: A straight-forward extension is unidentifiable, i.e. it is not possible to recover those latent components that actually generated the data. Here, we show that this problem can be solved by using temporal structure. We formulate two generative models in which the data is an arbitrary but invertible nonlinear transformation of time series (components) which are statistically independent of each other.

Artificial intelligence has held a place in our imagination from the beginning of the XX century. Already in the 1930s and 1940s, the pioneers of computing such as Alan Turing began formulating the basic techniques like neural networks that make today's AI possible. Today, AI is already all around us. Google uses Machine Learning to filter out spam messages from Gmail. Facebook trained computers to identify specific human faces nearly as accurately as humans do.

A few months ago I wrote a tutorial on how to classify images using Convolutional Neural Networks (specifically, VGG16) pre-trained on the ImageNet dataset with Python and the Keras deep learning library. The pre-trained networks inside of Keras are capable of recognizing 1,000 different object categories, similar to objects we encounter in our day-to-day lives with high accuracy. Back then, the pre-trained ImageNet models were separate from the core Keras library, requiring us to clone a free-standing GitHub repo and then manually copy the code into our projects. This solution worked well enough; however, since my original blog post was published, the pre-trained networks (VGG16, VGG19, ResNet50, Inception V3, and Xception) have been fully integrated into the Keras core (no need to clone down a separate repo anymore) -- these implementations can be found inside the applications sub-module. Because of this, I've decided to create a new, updated tutorial that demonstrates how to utilize these state-of-the-art networks in your own classification projects.

Machine learning couldn't be hotter. A type of artificial intelligence that enables computers to learn to perform tasks and make predictions without explicit programming, machine learning has caught fire among the hip tech set, but remains a somewhat futuristic concept for most enterprises. But thanks to technological advances and emerging frameworks, machine learning may soon hit the mainstream. Consulting firm Deloitte expects to see a big increase in the use and adoption of machine learning in the coming year. This is in large part because the technology is becoming much more pervasive.

We propose a technique for producing "visual explanations" for decisions from a large class of CNN-based models, making them more transparent. Our approach - Gradient-weighted Class Activation Mapping (Grad-CAM), uses the gradients of any target concept, flowing into the final convolutional layer to produce a coarse localization map highlighting the important regions in the image for predicting the concept. Unlike previous approaches, GradCAM is applicable to a wide variety of CNN model-families: (1) CNNs with fully-connected layers (e.g. VGG), (2) CNNs used for structured outputs (e.g. captioning), (3) CNNs used in tasks with multimodal inputs (e.g. VQA) or reinforcement learning, without any architectural changes or re-training. We combine GradCAM with fine-grained visualizations to create a high-resolution class-discriminative visualization and apply it to off-the-shelf image classification, captioning, and visual question answering (VQA) models, including ResNet-based architectures. In the context of image classification models, our visualizations (a) lend insights into their failure modes (showing that seemingly unreasonable predictions have reasonable explanations), (b) are robust to adversarial images, (c) outperform previous methods on weakly-supervised localization, (d) are more faithful to the underlying model and (e) help achieve generalization by identifying dataset bias. For captioning and VQA, our visualizations show that even non-attention based models can localize inputs. Finally, we conduct human studies to measure if GradCAM explanations help users establish trust in predictions from deep networks and show that GradCAM helps untrained users successfully discern a "stronger" deep network from a "weaker" one. Our code is available at https://github.com/ramprs/grad-cam. A demo and a video of the demo can be found at http://gradcam.cloudcv.org and youtu.be/COjUB9Izk6E.

We investigate different strategies for active learning with Bayesian deep neural networks. We focus our analysis on scenarios where new, unlabeled data is obtained episodically, such as commonly encountered in mobile robotics applications. An evaluation of different strategies for acquisition, updating, and final training on the CIFAR-10 dataset shows that incremental network updates with final training on the accumulated acquisition set are essential for best performance, while limiting the amount of required human labeling labor.

A cognitive business takes advantage of recent developments in cognitive computing to improve the overall effectiveness of its people, processes and technology. Data is starting to be pulled from more and more sources today to help solve problems in diverse fields – from health care to national defense and from daily operations to setting the right metrics that measure progress towards strategic and tactical goals. In 2016, the amount of global data being collected and analyzed is unprecedented--and growing. Working with that data in smarter ways is the key to future business success. For example, IBM's Watson relies on deep learning algorithms and neural networks to process information by comparing it to a teaching set of data in research hospitals to diagnose symptoms and recommend better patient treatment plans. As Internet of Things (IoT) sensors expand to new areas, and as artificial intelligence becomes more and more of a reality over the next two decades, businesses will rely on a new mix of real-time data, analytical processing and cutting edge alternative solutions to transform the way services are delivered. The right data will be collected and processed faster--yielding improved results for clients. Bottom line: a cognitive business is an increasingly smarter, data-driven business.

TLDR: In this blogpost, we're going to train a neural network that is fully encrypted during training (trained on unencrypted data). The result will be a neural network with two beneficial properties. First, the neural network's intelligence is protected from those who might want to steal it, allowing valuable AIs to be trained in insecure environments without risking theft of their intelligence. Secondly, the network can only make encrypted predictions (which presumably have no impact on the outside world because the outside world cannot understand the predictions without a secret key). This creates a valuable power imbalance between a user and a superintelligence. If the AI is homomorphically encrypted, then from it's perspective, the entire outside world is also homomorphically encrypted.

Your consumers spend a lot of time exploring and analyzing suitable information―which books to study, which news articles to read, which songs to play, which movies to watch, which games to play, and so on. Imagine, what their experience would be like, if they don't need to pick anything on their own, but are presented with options of their liking―be it in education or media or entertainment. Here are some of the things they can be offered: • Adaptive text-books, in which content changes based on the pace of learning and comfort level of the reader. Such advancements reduce the overall time spent on information discovery, and increase the scope of effective information consumption (or learning). Domains such as education, publishing, entertainment, and advertisement mostly deal with granular digital assets (text, images, audio, video, multi-media, and so on), and are better prepared to enhance personalization even without creating new content from scratch.