|February 5 · Issue #73 · View online |
Hey and welcome,
This weeks sponsor is O'Reilly Media who will be hosting The AI Conference in April.
Deep Learning Weekly subscribers can save 20% with the discount code DLWEEKLY20.
We’re proud to introduce a new blog on deeplearningjobs.com; Deep Learning in the Wild, where we interview companies and startups across industries to find out how deep learning is actually applied to tackle real-world problems:
| i2x̅ - Conversation training and analysis |
Ilya Edrenkin, head of machine learning at i2x̅, shares some insights into the work he’s doing.
| Lobster - Real-time stock photos |
Lobster Media gives insights on how they apply deep learning on stock photography and what challenges they had to tackle.
| People + AI Research - Library - Google Design |
Machine learning is changing how we build experiences and interact with the world. This collection shares practical insights from Google’s People + AI Research team on how to take a multidisciplinary and human-centered approach to designing with ML and AI.
| Google is launching an AI research center in France and expanding its office |
Google CEO Sundar Pichai wrote a blog post about Google’s investments in France. In addition to growing Google’s current teams, the company is going to create a new research center dedicated to artificial intelligence.
| A Guide to Receptive Field Arithmetic for Convolutional Neural Networks |
This post explains the concepts of a receptive field which is perhaps one of the most important concepts in Convolutional Neural Networks (CNNs) that deserves more attention from the literature.
| Annotating Large Datasets with the TensorFlow Object Detection API |
Learn how to use the TensorFlow Object Detection API to predict annotations for large datasets.
| A Refresher on Batch (re-)Normalization |
This post explains how batch renormalization deals with the problem that batch normalization can sometimes impair model performance and how renormalization works to remedy these defects. Highly recommended especially if you have been struggling understanding the concept of ‘internal covariate shift’.
| Automatically Test Neural Network Models in one Function Call |
| Building a simple Keras + Deep Learning REST API |
In this tutorial you will learn a simple method to take a Keras model and deploy it as a REST API.
The examples covered in this post will serve as a template/starting point for building your own deep learning APIs and the code can be extended and customized based on how scalable and robust your API endpoint needs to be.
| DeepMind: Learning Explanatory Rules from Noisy Data |
This blog posts expands on a recent DeepMind paper which demonstrates it is possible for systems to combine intuitive perceptual with conceptual interpretable reasoning. The system we describe, ∂ILP, is robust to noise, data-efficient, and produces interpretable rules.
| How to build your own AlphaZero AI using Python and Keras |
| Google Developers Blog: Announcing TensorFlow 1.5 |
Most notably, “Eager Execution” for TensorFlow is now available as a preview. This allows you to execute TensorFlow operations immediately as they are called from Python.
| FAIR: Countering Adversarial Image using Input Transformations. |
This package implements the experiments described in the paper Countering Adversarial Images Using Input Transformations
(see below). It contains implementations for adversarial attacks, defenses based image transformations, training, and testing convolutional networks under adversarial attacks using our defenses.
| CakeChat: Emotional Generative Dialog System |
CakeChat is a dialog system that is able to express emotions in a text conversation. It is written in Theano and Lasagne and uses end-to-end trained embeddings of 5 different emotions to generate responses conditioned by a given emotion. There is also a Wired article
about emotional chat bots.
| Autonomous Driving Cookbook |
In this tutorial by Microsoft walks you through how to train and test an end-to-end deep learning model for autonomous driving using data collected from the AirSim simulation environment.
The AI Conference is where cutting-edge science meets new business implementation. Presented by O'Reilly Media and Intel. Save 20% on most passes when you use the discount code DLWEEKLY20.
| Machine Learning Conferences in 2018 |
A comprehensive list of over 200 conferences on machine learning and AI taking place in 2018.
| Countering Adversarial Images using Input Transformations |
This paper investigates strategies that defend against adversarial-example attacks on image-classification systems by transforming the inputs before feeding them to the system. Specifically, the authors investigate applying image transformations such as bit-depth reduction, JPEG compression, total variance minimization, and image quilting before feeding the image to a convolutional network classifier.
| tempoGAN: A Temporally Coherent, Volumetric GAN for Super-resolution Fluid Flow |
The authors propose a temporally coherent generative model addressing the super-resolution problem for fluid flows. This work represents a first approach to synthesize four-dimensional physics fields with neural networks. Based on a conditional generative adversarial network that is designed for the inference of three-dimensional volumetric data, the proposed model generates consistent and detailed results by using a novel temporal discriminator, in addition to the commonly used spatial one.