Over the past few years, advances in deep learning have driven tremendous progress in image processing, speech recognition, and forecasting.

Welcome back to Mind Over Money. Since I am an investor in an exciting technology company you may have heard of called NVIDIA (NVDA), I often find myself in the position of having to explain to my followers and fellow investors "what exactly is AI" in a practical, right-now sense, and not some science fiction sense. NVDA's type of computer chip, the GPU, is at the heart of modern AI R&D and they sell a lot of them not just for advanced gaming graphics but also to industry for applications in autonomous driving where Tesla (TSLA), Toyota and Mercedes are customers. NVDA also has a bigger business selling their processors to big cloud companies like Amazon, Google (GOOGL), Microsoft, IBM (IBM) and Alibaba (BABA). If you want to learn more about NVDA's GPU chip technologies, see my video Get Your MPA in Deep Learning.

Anticipating user & business needs to alert & advise logical steps to increase efficiency. What impact will predictive intelligence have on business efficiency & personal organization?

Chida Chidambaram Vishal Deshpande BDT311 Deep Learning Going Beyond Machine Learning October 2015 2. What to Expect from the Session Data analytics options on AWS Machine learning (ML) – high level Amazon ML from AWS ML sample use case Deep learning (DL) – high level DL sample use cases AWS GPU/HPCC server family Q&A 3. Data Analytics Options on AWS Amazon EMR AnalyzeStoreIngest Amazon Kinesis DynamoDB Amazon Redshift RDSS3 Amazon Kinesis Consumer Machine Learning Amazon Kinesis Producer Traditional Server Mobile Clients EC2 Machines 5. Machine Learning How can a machine identify Bruce Willis vs Jason Statham? Bruce Willis??? 6. Machine Learning Machine Learning Artificial Intelligence Optimization & Control Neuroscience and Neural Networks Statistical Modeling Information Theory 7. Machine Learning Bear Eagle People Sunset 8. Machine Learning • Using machines to discover trends and patterns and compute mathematical predictive models based on factual past data • ML models provide insights into likely outcomes based on the past – machine learning helps uncover the probability of an outcome in the future rather than merely state what has already happened in the past • Past data and statistical modeling is used to make predictions based on probability Where traditional business analytics aims at answering questions about past events, machine learning aims at answering questions about the possibilities of future events 9. Machine Learning Supervised learning Human intervention and validation required Photo classification and tagging Unsupervised learning No human intervention required Auto-classification of documents based on context 10. Machine Learning – Process How can a machine identify Bruce Willis vs Jason Statham? Image analysis – Input feature set for image 1 - bald, black suit Bruce Willis??? 14. Machine Learning – Process • Start with data for which the answer is already known • Identify the target – what you want to predict from the data • Pick the variables/features that can be used to identify the patterns to predict the target • Train the ML model with the dataset for which you already know the target answer • Use the trained model to predict the target on the data for which the answer is not known • Evaluate the model for accuracy • Improve the model accuracy as needed 15. Machine Learning – When to Use It You need ML if • Simple classification rules are inadequate • Scalability is an issue with large number of datasets You do not need ML if • You can predict the answers by using simple rules and computations • You can program predetermined steps without needing any data driven learning 16.

There have been many "ages" throughout human history, most notably the industrial age and the digital age. Now, we have officially entered the age of artificial intelligence (AI). Within this AI age are many technologies, including machine learning and deep learning. These are fundamentally transforming and altering the business landscape. Its ability to revolutionize the world has been likened to what electricity did in its day.

Deep neural networks (DNNs) have become increasingly important due to their excellent empirical performance on a wide range of problems. However, regularization is generally achieved by indirect means, largely due to the complex set of functions defined by a network and the difficulty in measuring function complexity. There exists no method in the literature for additive regularization based on a norm of the function, as is classically considered in statistical learning theory. In this work, we propose sampling-based approximations to weighted function norms as regularizers for deep neural networks. We provide, to the best of our knowledge, the first proof in the literature of the NP-hardness of computing function norms of DNNs, motivating the necessity of a stochastic optimization strategy. Based on our proposed regularization scheme, stability-based bounds yield a $\mathcal{O}(N^{-\frac{1}{2}})$ generalization error for our proposed regularizer when applied to convex function sets. We demonstrate broad conditions for the convergence of stochastic gradient descent on our objective, including for non-convex function sets such as those defined by DNNs. Finally, we empirically validate the improved performance of the proposed regularization strategy for both convex function sets as well as DNNs on real-world classification and segmentation tasks.

Learning social media data embedding by deep models has attracted extensive research interest as well as boomed a lot of applications, such as link prediction, classification, and cross-modal search. However, for social images which contain both link information and multimodal contents (e.g., text description, and visual content), simply employing the embedding learnt from network structure or data content results in sub-optimal social image representation. In this paper, we propose a novel social image embedding approach called Deep Multimodal Attention Networks (DMAN), which employs a deep model to jointly embed multimodal contents and link information. Specifically, to effectively capture the correlations between multimodal contents, we propose a multimodal attention network to encode the fine-granularity relation between image regions and textual words. To leverage the network structure for embedding learning, a novel Siamese-Triplet neural network is proposed to model the links among images. With the joint deep model, the learnt embedding can capture both the multimodal contents and the nonlinear network information. Extensive experiments are conducted to investigate the effectiveness of our approach in the applications of multi-label classification and cross-modal search. Compared to state-of-the-art image embeddings, our proposed DMAN achieves significant improvement in the tasks of multi-label classification and cross-modal search.

We develop an end-to-end training algorithm for whole-image breast cancer diagnosis based on mammograms. It requires lesion annotations only at the first stage of training. After that, a whole image classifier can be trained using only image level labels. This greatly reduced the reliance on lesion annotations. Our approach is implemented using an all convolutional design that is simple yet provides superior performance in comparison with the previous methods. On DDSM, our best single-model achieves a per-image AUC score of 0.88 and three-model averaging increases the score to 0.91. On INbreast, our best single-model achieves a per-image AUC score of 0.96. Using DDSM as benchmark, our models compare favorably with the current state-of-the-art. We also demonstrate that a whole image model trained on DDSM can be easily transferred to INbreast without using its lesion annotations and using only a small amount of training data.

Intel has a reputation of making fast chips, but none are very efficient at the most trending thing in computing right now – Artificial intelligence (AI). Deep-learning apps compatible with computer vision, voice recognition and other tasks most of the times need to run matrix calculations on gigantic arrays -- something that doesn't suit general-purpose Core or Xeon chips. However, with the purchase of deep learning chipmaker Nervana, Intel will ship its first purpose-built AI chips, by the end of 2017. Brian Krzanich, Intel CEO during his keynote speech at WSJDLive shared that close collaboration with Facebook and their technical insights have enable them to bring this new generation of AI hardware, the Nervana Neural Processor family (NNP), to market. Along with social media, Intel is targeting other applications such as weather, automotive, and healthcare.

Starting with the Google DeepMind paper, there has been a lot of new attention around training models to play video games. You, the data scientist/engineer/enthusiast, may not work in reinforcement learning but probably are interested in teaching neural networks to play video games. The lessons below were gleaned from working on my own implementation of the Nature paper. The lessons are aimed at people who work with data but may run into some issues with some of the non-standard approaches used in the reinforcement learning community when compared with typical supervised learning use cases. I will address both technical details of the parameters of the neural networks and the libraries involved.