Canonical transformation plays a fundamental role in simplifying and solving classical Hamiltonian systems. We construct flexible and powerful canonical transformations as generative models using symplectic neural networks. The model transforms physical variables towards a latent representation with an independent harmonic oscillator Hamiltonian. Correspondingly, the phase space density of the physical system flows towards a factorized Gaussian distribution in the latent space. Since the canonical transformation preserves the Hamiltonian evolution, the model captures nonlinear collective modes in the learned latent representation. We present an efficient implementation of symplectic neural coordinate transformations and two ways to train the model. The variational free energy calculation is based on the analytical form of physical Hamiltonian. While the phase space density estimation only requires samples in the coordinate space for separable Hamiltonians. We demonstrate appealing features of neural canonical transformation using toy problems including two-dimensional ring potential and harmonic chain. Finally, we apply the approach to real-world problems such as identifying slow collective modes in alanine dipeptide and conceptual compression of the MNIST dataset.

M uller, 3 S oren Schleibaum, 3 Mark V ollrath 5 1 Department of Computer Science, Bar-Ilan University, Israel (email: sarit@cs.biu.ac.il) 2 Department of Computer Science, Ariel University, Israel 3 Department of Informatics, TU Clausthal, Germany 4 Chair of Information Management, Georg-August-Universitat G ottingen, Germany 5 Chair of Engineering and Traffic Psychology, TU Braunschweig, Germany Abstract Explanation is necessary for humans to understand and accept decisions made by an AI system when the system's goal is known. It is even more important when the AI system makes decisions in multi-agent environments where the human does not know the systems' goals since they may depend on other agents' preferences. In such situations, explanations should aim to increase user satisfaction, taking into account the system's decision, the user's and the other agents' preferences, the environment settings and properties such as fairness, envy and privacy. Generating explanations that will increase user satisfaction is very challenging; to this end, we propose a new research direction: Explainable decisions in Multi-Agent Environments (xMASE). We then review the state of the art and discuss research directions towards efficient methodologies and algorithms for generating explanations that will increase users' satisfaction from AI system's decisions in multi-agent environments. Introduction Many AI systems need to make decisions in multi-agent environments where the agents, including people and robots, have possibly conflicting preferences. The system should balance between these preferences when making decisions regarding all agents.

OpenCV is today's most popular image processing library, covering everything from classic image processing algorithms to cutting-edge deep learning pretrained models. However because OpenCV is not differentiable it mainly focuses on pre-processing tasks and cannot be embedded in an entire training process. Inspired by OpenCV, Kornia is based on PyTorch and designed to solve generic computer vision problems. It contains a set of routines for performing color space conversions, and differentiable modules for performing tasks such as image filtering and edge detection. Kornia's core code can efficiently define and compute the gradient of complex functions with reverse-mode auto-differentiation. Kornia consists of subset packages containing operators which can be inserted into neural networks to enable models to perform tasks such as image transformations, epipolar geometry, and depth estimation.

With its ability to sift through large amounts of data, AI has the potential to make medical examinations more precise, reliable, and efficient. For example, in CT-based lung cancer screening, researchers at Philips have demonstrated that a deep learning algorithm may be helpful to radiologists as a decision support tool or second opinion. Similarly, in digital pathology, algorithms can point to regions of interest in tissue samples that demand further inspection by the pathologist, while making it easier to discard slides without signs of cancer. And by unravelling the molecular mechanisms that give rise to an individual's cancer, we can truly begin to understand how targeted treatment may help that individual. Next to the interpretation of medical data, AI can also support with ancillary tasks such as patient scheduling.

A key avenue for deploying deep learning models is a mobile device. The advantage of running models in mobile apps instead of sending them to the cloud is the reduction in latency and the ability to ensure data privacy for users. Despite the variety of deep learning libraries and AI tools, successfully embedding a deep learning model into a mobile app can be challenging. Anirudh Koul, head of AI and research at Aira explained the step by step process for deploying convolutional neural networks (ConvNets) in applications for mobile devices during his lecture at ODSC West 2018. Anirudh explained the process necessary to deploy a ConvNet application given a few time frames.

Deep neural networks excel at image recognition, but are easily hacked. A self-driving car approaches a stop sign, but instead of slowing down, it accelerates into the busy intersection. An accident report later reveals that four small rectangles had been stuck to the face of the sign. These fooled the car's onboard artificial intelligence (AI) into misreading the word'stop' as'speed limit 45'. Such an event hasn't actually happened, but the potential for sabotaging AI is very real.

Facebook today introduced Captum, a library for explaining decisions made by neural networks with deep learning framework PyTorch. Captum is designed to implement state of the art versions of AI models like Integrated Gradients, DeepLIFT, and Conductance. Captum allows researchers and developers to interpret decisions made in multimodal environments that combine, for example, text, images, and video, and allows them to compare results to existing models within the library. Developers can also use Captum to understand feature importance or perform a deep dive on neural networks to understand neuron and layer attributions. The tool will also launch with Captum Insights, a visualization tool for visual representations of Captum results.

Monitoring the right GPU performance metrics can go a long way in helping you train and deploy deep learning applications. GPU utilization is one of the primary metrics to observe during a deep learning training session. This metric is readily accessible through popular GPU monitoring interfaces such as NVIDIA's "NVIDIA-smi". A GPU's utilization is defined as the percentage of time one or more GPU kernels are running over the last second, which is analogous to a GPU being utilized by a deep learning program. Monitoring your deep learning training sessions' GPU utilization is one of the best indicators to determine if your GPU is actually being used. Moreover, monitoring the real-time utilization trend can help identify bottlenecks in your pre-processing and feature engineering pipelines that might be slowing down your training process.

Almost every day we read about companies and their Artificial Intelligence (AI) strategies. Sometimes it feels like an arms race where businesses feel they will get left behind if they can't claim to have AI and (usually) deep learning embedded somewhere in their product. We have seen this type of thing before, reminiscent of the social media and big data hypes of years gone by. It used to be that companies were tripping over themselves to be seen as "big data" now the focus is on "AI " as they try to position themselves as appealing to customers and investors – one report estimates as much as 40% of European startups classified as "AI" don't actually use AI in any material way! The hype around AI has largely been driven by substantial recent progress in the sub field of deep learning.

On Thursday the developers of PyTorch announced PyTorch Mobile, which they say will allow for "end-to-end workflow from Python to deployment on iOS and Android." PyTorch Mobile is part of PyTorch 1.3, which currently is an "experimental release" that the organization will be "building on over the next couple of months." PyTorch 1.2 was released in August. New features coming will include preprocessing and integration APIs, support for ARM CPUs and QNNPACK (a quantized neural network package designed for PyTorch), build-level optimization, and performance enhancements for mobile CPUs/GPUs. Android builds will use the Maven plug-in and iOS will use CocoaPods with Swift.