We present a neural sequence model designed specifically for symbolic music. The model is based on a learned edit distance mechanism which generalises a classic recursion from computer science, leading to a neural dynamic program. Repeated motifs are detected by learning the transformations between them. We represent the arising computational dependencies using a novel data structure, the edit tree; this perspective suggests natural approximations which afford the scaling up of our otherwise cubic time algorithm. We demonstrate our model on real and synthetic data; in all cases it outperforms a strong stacked long short-term memory benchmark.

"For instance, on the planet Earth, man had always assumed that he was more intelligent than dolphins because he had achieved so much--the wheel, New York, wars and so on--whilst all the dolphins had ever done was muck about in the water having a good time. But conversely, the dolphins had always believed that they were far more intelligent than man--for precisely the same reasons." Douglas Adams, The Hitchhiker's Guide to the Galaxy One of the big obstacles in the field of artificial intelligence is not having a definition of intelligence based on solid mathematical and physical principles that could inspire the design and implementations of efficient intelligent algorithms. For instance, consider the most widely accepted definition of intelligence, signed by 52 specialist on the field [2]: "A very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience. It is not merely book learning, a narrow academic skill, or test-taking smarts. Rather, it reflects a broader and deeper capability for comprehending our surroundings..." A more recent definition [3] provided by Shane Legg, chief scientist of Deep Mind, and Marcus Hutter, founder of AIXI, is the following: "Intelligence measures an agent's ability to achieve goals in a wide range of environments." Although there are many other definitions of intelligence, they are too fuzzy to help us develop a theory of intelligent behaviour or give us an insight on how a general, computable and efficient algorithm for generating intelligent behaviour should look like. This document is an effort to present such a definition based on entropic principles deeply inspired by the concept of "Causal Entropic Forces" introduced by Alexander Wissner-Gross in 2013 [1] and to propose a generic implementation of those principles.

We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a blackbox differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models.

Designing models that are robust to small adversarial perturbations of their inputs has proven remarkably difficult. In this work we show that the reverse problem---making models more vulnerable---is surprisingly easy. After presenting some proofs of concept on MNIST, we introduce a generic tilting attack that injects vulnerabilities into the linear layers of pre-trained networks without affecting their performance on natural data. We illustrate this attack on a multilayer perceptron trained on SVHN and use it to design a stand-alone adversarial module which we call a steganogram decoder. Finally, we show on CIFAR-10 that a state-of-the-art network can be trained to misclassify images in the presence of imperceptible backdoor signals. These different results suggest that adversarial perturbations are not always informative of the true features used by a model.

The robotic systems continuously interact with complex dynamical systems in the physical world. Reliable predictions of spatiotemporal evolution of these dynamical systems, with limited knowledge of system dynamics, are crucial for autonomous operation. In this paper, we present HybridNet, a framework that integrates data-driven deep learning and model-driven computation to reliably predict spatiotemporal evolution of a dynamical systems even with in-exact knowledge of their parameters. A data-driven deep neural network (DNN) with Convolutional LSTM (ConvLSTM) as the backbone is employed to predict the time-varying evolution of the external forces/perturbations. On the other hand, the model-driven computation is performed using Cellular Neural Network (CeNN), a neuro-inspired algorithm to model dynamical systems defined by coupled partial differential equations (PDEs). CeNN converts the intricate numerical computation into a series of convolution operations, enabling a trainable PDE solver. With a feedback control loop, HybridNet can learn the physical parameters governing the system's dynamics in real-time, and accordingly adapt the computation models to enhance prediction accuracy for time-evolving dynamical systems. The experimental results on two dynamical systems, namely, heat convection-diffusion system, and fluid dynamical system, demonstrate that the HybridNet produces higher accuracy than the state-of-the-art deep learning based approach.

Figure 1: A single robot grows calluses as it walks, in response to pressure on its feet (youtu.be/0cmwpcxSUWI). Typically, AI researchers and roboticists try to realize intelligent behavior in machines by tuning parameters of a predefined structure (body plan and/or neural network architecture) using evolutionary or learning algorithms. Another but not unrelated longstanding property of these systems is their brittleness to slight aberrations, as highlighted by the growing deep learning literature on adversarial examples. Here we show robustness can be achieved by evolving the geometry of soft robots, their control systems, and how their material properties develop in response to one particular interoceptive stimulus (engineering stress) during their lifetimes. By doing so we realized robots that were equally fit but more robust to extreme material defects (such as might occur during fabrication or by damage thereafter) than robots that did not develop during their lifetimes, or developed in response to a different interoceptive stimulus (pressure). This suggests that the interplay between changes in the containing systems of agents (body plan and/or neural architecture) at different temporal scales (evolutionary and developmental) along different modalities (geometry, material properties, synaptic weights) and in response to different signals (interoceptive and external perception) all dictate those agents' abilities to evolve or learn capable and robust strategies. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page.

Man has always been fascinated with a view of the world from the top -- building watch-towers, high fortwalls, capturing the highest mountain peak. To capture a glimpse and share it with the world, people went to great lengths to defy gravity, enlisting the help of ladders, tall buildings, kites, balloons, planes, and rockets. Today, access to drones that can fly as high as 2kms is possible even for the general public. These drones have high resolution cameras attached to them that are capable of acquiring quality images which can be used for various kinds of analysis. With easier access to drones, we're seeing a lot of interest and activity by photographers & hobbyists, who are using it to make creative projects such as capturing inequality in South Africa or breathtaking views of New York which might make Woody Allen proud.

In a bid to further strengthen its fintech portfolio, digital payments firm MobiKwik today announced a strategic investment of Rs. 2 crores in Pune based data science company, Pivotchain Solutions. Founded in February 2017 by Deepak Rao and Yogendra Pratap Singh, Pivotchain is a Predictive Analytics company with expertise in Machine Learning & Artificial Intelligence. It has built exclusive AI and deep learning models for Mobikwik. These models will be instrumental for MobiKwik as it rolls out lending products to address the credit requirements of its user base. Speaking on the strategic investment, Bipin Preet Singh, Founder and CEO, MobiKwik, said, "MobiKwik is transforming from a leading digital payments player, to India's largest digital financial services platform. Delivering high quality fintech products will require immense focus on data, and an in-depth understanding of the user requirements, across categories. Pivotchain is doing incredible work in alternate data scoring, predictive modeling & risk management and this investment will give us an edge over competition. We will continue to invest in companies that can add value to our business."

Neurala, Inc. developed The Neurala Brain--deep learning neural network software that makes devices and products like drones, mobile phones and cameras more intelligent, engaging and useful. Neurala provides customized solutions ranging from high-end applications to inexpensive everyday consumer products. With The Neurala Brain and an ordinary camera, products can learn people and objects, recognize them in a video stream, find them in the video, and track them as they move. The Neurala Brain is based on technology originally developed for NASA and the U.S. Air Force. Follow Neurala on Twitter @Neurala and on Facebook, YouTube and LinkedIn.

I recently caught up with Dr. Bhushan Desam, AI global business leader for Lenovo's Data Center Group to discuss how the digital transformation of business isn't truly possible without incorporating machine learning. As a global business leader, Bhushan is focused on developing AI and machine learning business at DCG. On any given day, Bhushan helps manage Lenovo's overall AI strategy, assists in making product portfolio decisions with engineers and product teams, interfaces with R&T teams, supports global sales teams with customer engagement and deepens relationships with HPC customers and partners. He completed his PhD in engineering at the University of Utah, which introduced Bhushan to high performance computing (HPC) and laid the technical groundwork for his role and responsibilities today. After several years working in the field as a researcher and an engineer, Bhushan re-entered academia in 2011 to pursue a joint program in technology management at MIT Sloan School of Management and School of Engineering.