TensorFlow is a popular framework used for machine learning. The Amazon Deep Learning AMI comes bundled with everything you need to start using TensorFlow from development through to production. In this Lab, you will develop, visualize, serve, and consume a TensorFlow machine learning model using the Amazon Deep Learning AMI.

Predictive analytics and automation--through AI and machine learning--are increasingly being integrated into enterprise applications to support decision making and address critical issues such as security and business intelligence. Public cloud platforms like AWS offer dedicated services that allow companies to easily implement deep learning models, sometimes even without requiring specialized skills in data modeling or analytics. TensorFlow is an open-source, powerful framework for developing machine learning models.It is one of the most popular machine learning frameworks because it is flexible and it caters to different levels of data science knowledge. The Amazon Deep Learning AMI comes pre-configured with everything you need to start using TensorFlow from development to production. It allows teams to get started training quickly without worrying about dependencies or costly installations. Eager to try it out?

Appier recently showcased its pioneering research that demonstrated AI's ability to be creative, at the 32nd AAAI Conference on Artificial Intelligence (AAAI-18). AI's ability to create and design items that humans will find appealing is considered the next important breakthrough in the field - and Appier's research suggests that the day when AI can help designers in their day-to-day work might be closer than we thought. The Appier team, comprising Yong-Siang Shih, Dr Kai-Yueh Chang and Dr Hsuan-Tien Lin from Appier, and Professor Min Sun from National Tsing Hua University, designed an AI model that first mimics the human ability to put together and design outfits, after observing and learning from examples of good design. They did this with deep learning techniques in combination with a new Projected Compatibility Distance (PCD) methodology which they designed, that discerns the distance between compatible items. This compatibility distance is then used to run through a unique application of Generative Adversarial Networks (GANs) to generate novel compatible apparel items when apparel images are input.

Intel started a new initiative to help developers, that are working on Artificial Intelligence-based products, to bring their devices to market. The new program will make the process easy for developers and enable them to launch their prototypes in the market. "Once developers have a prototype, the next step is to take it into production, which can be challenging and costly for small companies and entrepreneurs. To make it easier, Intel selected AAEON Technologies, a leading manufacturer of advanced industrial and embedded computing platforms, as the first Intel AI: In Production partner. Through the program, AAEON provides two streamlined production paths for developers integrating the low-power Intel Movidius Myriad 2 Vision Processing Unit (VPU) into their product designs."

"No one wants to be sold but everyone wants to buy." Most of us hate being sold. The moment we know someone is selling something, we keep our guards up. In the book, The Challenger Sale, authors Mathew Dixon and Brent Adamson surveyed over 6000 salespeople from around the world and found that'challenger salespeople' outperformed every other group. Who are these challenger salespeople? These are the people who challenge the norm, are more knowledgeable and educate their customers.

Somewhere in a remote town at an indoor vertical farm, sensors and cameras are installed to gather data on growing environment conditions; capturing everything from the moisture and nutrients to the availability of light and oxygen. This information is then sent to the processors through the cloud to be analysed and to suggest immediate action that is performed under strict vigilance. In another instance, data scientists are mining satellite images of parking lots of shopping malls to predict foot traffic for major retailing corporations. Taking things a notch higher, deep learning algorithms are applied to get data for investors to buy, who then make stock allocation decisions based on this information. That is the future of thinking and the wonder of data science.

The use of back-propagation and its variants to train deep networks is often problematic for new users, with issues such as exploding gradients, vanishing gradients, and high sensitivity to weight initialization strategies often making networks difficult to train. In this paper, we present Local Representation Alignment (LRA), a training procedure that is much less sensitive to bad initializations, does not require modifications to the network architecture, and can be adapted to networks with highly nonlinear and discrete-valued activation functions. Furthermore, we show that one variation of LRA can start with a null initialization of network weights and still successfully train networks with a wide variety of nonlinearities, including tanh, ReLU-6, softplus, signum and others that are more biologically plausible. Experiments on MNIST and Fashion MNIST validate the performance of the algorithm and show that LRA can train networks robustly and effectively, succeeding even when back-propagation fails and outperforming other alternative learning algorithms, such as target propagation and feedback alignment.

We investigate the effects of initialization and architecture on the start of training in deep ReLU nets. We identify two common failure modes for early training in which the mean and variance of activations are poorly behaved. For each failure mode, we give a rigorous proof of when it occurs at initialization and how to avoid it. The first failure mode, exploding/vanishing mean activation length, can be avoided by initializing weights from a symmetric distribution with variance 2/fan-in. The second failure mode, exponentially large variance of activation length, can be avoided by keeping constant the sum of the reciprocals of layer widths. We demonstrate empirically the effectiveness of our theoretical results in predicting when networks are able to start training. In particular, we note that many popular initializations fail our criteria, whereas correct initialization and architecture allows much deeper networks to be trained.

Deploying state-of-the-art CNNs requires power-hungry processors and off-chip memory. This precludes the implementation of CNNs in low-power embedded systems. Recent research shows CNNs sustain extreme quantization, binarizing their weights and intermediate feature maps, thereby saving 8-32 memory and collapsing energy-intensive sum-of-products into XNOR-and-popcount operations. We present XNORBIN, an accelerator for binary CNNs with computation tightly coupled to memory for aggressive data reuse. Implemented in UMC 65nm technology XNORBIN achieves an energy efficiency of 95 TOp/s/W and an area efficiency of 2.0 TOp/s/MGE at 0.8 V. I.

We consider learning of submodular functions from data. These functions are important in machine learning and have a wide range of applications, e.g. data summarization, feature selection and active learning. Despite their combinatorial nature, submodular functions can be maximized approximately with strong theoretical guarantees in polynomial time. Typically, learning the submodular function and optimization of that function are treated separately, i.e. the function is first learned using a proxy objective and subsequently maximized. In contrast, we show how to perform learning and optimization jointly. By interpreting the output of greedy maximization algorithms as distributions over sequences of items and smoothening these distributions, we obtain a differentiable objective. In this way, we can differentiate through the maximization algorithms and optimize the model to work well with the optimization algorithm. We theoretically characterize the error made by our approach, yielding insights into the trade-off of smoothness and accuracy. We demonstrate the effectiveness of our approach for jointly learning and optimizing on synthetic maxcut data, and on a real world product recommendation application.