Several factors have impacted the proliferation at which the research and development of AI is developing. The increase in computational resources, the explosive growth of data which stands at 48% year-on-year, based on Mary Meeker's 2017 Internet Trends Report, along with the decreasing cost of data storage, and the surge of open source frameworks, in addition to the shift from broad AI to industry-focused AI, have enabled Machine Learning and Deep Learning to therefore accelerate the evolvement of AI.

In case the text is composed of speech, speech-to-text conversion is performed.

We have taken a look at some of the challenges to overcome and predictions for its implementation from experts in the field who envision it becoming more practical and useful, automating some jobs and augmenting many others, combining machine learning and big data for fresh actionable insights. A deep learning system is, in short, a multi-layered neural network that learns representations of the world and stores them as a nested hierarchy of concepts many layers deep. For example, when processing thousands of images of human faces, it recognises objects based on a hierarchy of simpler building blocks: straight lines and curved lines at the basic level; then eyes, mouths, and noses; entire faces; and finally, specific facial features. Besides image recognition, deep learning offers the potential to approach complex challenges such as speech comprehension, human-machine conversation, language translation, and vehicle navigation, amongst others. How can we expect this technology to be implemented in the coming year?

The receptive field is perhaps one of the most important concepts in Convolutional Neural Networks (CNNs) that deserves more attention from the literature. All of the state-of-the-art object recognition methods design their model architectures around this idea. However, to my best knowledge, currently there is no complete guide on how to calculate and visualize the receptive field information of a CNN. This post fills in the gap by introducing a new way to visualize feature maps in a CNN that exposes the receptive field information, accompanied by a complete receptive field calculation that can be used for any CNN architecture. I've also implemented a simple program to demonstrate the calculation so that anyone can start computing the receptive field and gain better knowledge about the CNN architecture that they are working with.

Trace3 will play a key role in delivering NVIDIA-based artificial intelligence, machine learning, and deep learning solutions to enterprises worldwide. The NVIDIA partner program will include initiatives to help partners expand capabilities for the integration and deployment of NVIDIA GPU computing solutions, including NVIDIA DGX systems. The initial phase will focus on providing services for deep learning and neural network development for image analysis, natural language processing, and time-series analysis. Trace3 has been providing big data services to enterprise clients since the launch of their Data Intelligence practice in March 2014. Today, the company has new comprehensive Artificial Intelligence engagements (Server Hardware and Services) in the works at companies in the manufacturing, financial services, and banking industries as well as several other clients in the planning stages for 2018.

This code aims to reproduce results obtained in the paper "Visual Feature Attribution using Wasserstein GANs" This repository contains the code to reproduce results for the paper cited above, where the authors presents a novel feature attribution technique based on Wasserstein Generative Adversarial Networks (WGAN). The code works for both synthetic (2D) and real 3D neuroimaging data, you can check below for a brief description of the two datasets. Here is an example of what the generator/mapper network should produce: ctrl-click on the below image to open the gifv in a new tab (one frame every 50 iterations, left: input, right: anomaly map for synthetic data at iteration 50 * (its 1)). "Data: In order to quantitatively evaluate the performance of the examined visual attribution methods, we generated a synthetic dataset of 10000 112x112 images with two classes, which model a healthy control group (label 0) and a patient group (label 1). The images were split evenly across the two categories. We closely followed the synthetic data generation process described in [31][SubCMap: Subject and Condition Specific Effect Maps] where disease effects were studied in smaller cohorts of registered images. The control group (label 0) contained images with ran- dom iid Gaussian noise convolved with a Gaussian blurring filter. Examples are shown in Figure 1. The patient images (label 1) also contained the noise, but additionally exhib- ited one of two disease effects which was generated from a ground-truth effect map: a square in the centre and a square in the lower right (subtype A), or a square in the centre and a square in the upper left (subtype B). Importantly, both dis- ease subtypes shared the same label. The location of the off-centre squares was randomly offset in each direction by a maximum of 5 pixels. This moving effect was added to make the problem harder, but had no notable effect on the outcome."

We learn driving a car -- by driving.

A new paradigm is beginning to emerge in Radiology with the advent of increased computational capabilities and algorithms. This has led to the ability of real time learning by computer systems of different lesion types to help the radiologist in defining disease. For example, using a deep learning network, we developed and tested a multiparametric deep learning (MPDL) network for segmentation and classification using multiparametric magnetic resonance imaging (mpMRI) radiological images. The MPDL network was constructed from stacked sparse autoencoders with inputs from mpMRI. Evaluation of MPDL consisted of cross-validation, sensitivity, and specificity. Dice similarity between MPDL and post-DCE lesions were evaluated. We demonstrate high sensitivity and specificity for differentiation of malignant from benign lesions of 90% and 85% respectively with an AUC of 0.93. The Integrated MPDL method accurately segmented and classified different breast tissue from multiparametric breast MRI using deep leaning tissue signatures.

We study the sparse entropy-regularized reinforcement learning (ERL) problem in which the entropy term is a special form of the Tsallis entropy. The optimal policy of this formulation is sparse, i.e.,~at each state, it has non-zero probability for only a small number of actions. This addresses the main drawback of the standard Shannon entropy-regularized RL (soft ERL) formulation, in which the optimal policy is softmax, and thus, may assign a non-negligible probability mass to non-optimal actions. This problem is aggravated as the number of actions is increased. In this paper, we follow the work of Nachum et al. (2017) in the soft ERL setting, and propose a class of novel path consistency learning (PCL) algorithms, called {\em sparse PCL}, for the sparse ERL problem that can work with both on-policy and off-policy data. We first derive a {\em sparse consistency} equation that specifies a relationship between the optimal value function and policy of the sparse ERL along any system trajectory. Crucially, a weak form of the converse is also true, and we quantify the sub-optimality of a policy which satisfies sparse consistency, and show that as we increase the number of actions, this sub-optimality is better than that of the soft ERL optimal policy. We then use this result to derive the sparse PCL algorithms. We empirically compare sparse PCL with its soft counterpart, and show its advantage, especially in problems with a large number of actions.

We investigate the loss surface of deep linear and nonlinear neural networks. We show that for deep linear networks with differentiable losses, critical points after the multilinear parameterization inherit the structure of critical points of the underlying loss with linear parameterization. As corollaries we obtain "local minima are global" results that subsume most previous results, while showing how to distinguish global minima from saddle points. For nonlinear neural networks, we prove two theorems showing that even for networks with one hidden layer, there can be spurious local minima. Indeed, for piecewise linear nonnegative homogeneous activations (e.g., ReLU), we prove that for almost all practical datasets there exist infinitely many local minima that are not global. We conclude by constructing a counterexample involving other activation functions (e.g., sigmoid, tanh, arctan, etc.), for which there exists a local minimum strictly inferior to the global minimum.