Deep Learning in the Real World

Deep Learning in the Real World

Presented By The Open Data Science Conference
The Open Data  Science Conference
The Open Data Science Conference

Presentation Description

Deep Learning in the Real World

Speech recognition, image recognition, machine translation and autonomous machines are a few of the applications that have achieved incredible breakthroughs thanks to deep learning over the last 5 years.  As the models, platforms, and data sets improve deep learning will continue to be at the forefront of AI revolution.

Each ODSC conference host some of the world’s top experts in data science, including the field of deep learning.  Hear from 10 experts at some of the biggest names in the industry how deep learning is put to work in the real world. 


Talk Descriptions

Deep Learning and Language Processing at Airbnb - Avneesh Saluja, PHD, Machine Learning Scientist, AI Lab at AirBnB

Airbnb has had over 200 Million guest arrivals over all time, and over 40 Million in 2015 alone. Naturally, this exponential growth is a challenge to deal with from the perspective of the sheer amount of user-generated text created. Every listing contains text written by a host (listing descriptions) and text written by guests (reviews). Customer service issues come in over chat and email, and guests and hosts exchange hundreds of thousands of messages every day. This talk will present an overview of some of the work we have done in this area and will also preview a lot of the work that still needs to be done.

Deep Learning in the real world - Lukas Biewald, Founder of CrowdFlower

As companies take machine learning out of R&D and into production they face a whole set of new challenges. Human in the loop, active learning, and transfer learning are all essential design patterns for making deep learning real.

Deep Defense: Using Deep Learning to Fight Off Uber Fraudsters

Fraud models are generally based on narrow data streams processed by traditional machine learning models such as gradient boosted machines. Our talk will cover how Uber improved on this by applying deep learning to extract complex feature relationships from high-dimensional datasets such as tapstream and location data. We will cover the lessons we learned while applying deep learning to three fraud use cases:
Finding anomalous trip locations based on all Uber trip history, using tap streams to model normal vs fraud app usage and computer vision for validating credit cards and IDs"

Deep Learning with Tensorflow for Absolute Beginners

TensorFlow now provides the high-level API that let you write a few lines of Python code to get started with neural network, without understanding the hard math. Try this codelab to see how machine learning works on your laptop. This codelab is designed as an easy TensorFlow introduction for non ML experts. All you need to know is how to use Python. It would take about 2 - 3 hours to go through all the sections.

Deep Learning for Developers

Deep Learning has become the hottest topic in the IT industry. However, its arcane jargon and its intimidating equations often discourage software developers, who wrongly think that they’re “not smart enough.” In this session, we’ll explain the basic concepts of Neural Networks and Deep Learning in simple terms, with minimal theory and math. Then, through code-level demos based on Apache MXNet, we’ll demonstrate how to build, train and use models based on different types of networks: multi-layer perceptrons, convolutional neural networks and long short-term memory networks. Finally, we’ll share some optimisation tips which will help improve the training speed and the performance of your models.

Putting Deep Learning to Work - Alex Ermolaev, Head of AI at NVIDIA

The latest wave of AI innovation called “deep learning” caught many people by surprise. With rapidly improving accuracy and sophistication, the applications appear to be limitless. However, getting started with deep learning is not an easy task. In this talk, we are going to discuss the best practices for getting started with deep learning. We are also going to look at what makes successful target project for deep learning and what areas are best suited for other approaches. Finally, we are also going to look at several of the most common use cases, which are getting implemented by leading enterprises right now.

Deep Learning From Scratch Using Python - Seth Weidmad, Senior Data Scientist at Metis

Many of us have used libraries like Keras and TensorFlow to train Deep Learning models. But very few of us fully understand what is going on "under the hood." In this talk, we'll walk through how to create Deep Neural Networks powerful enough to solve complex image classification tasks, from scratch, using Python. We'll do everything from coding the layers of our network using classes to implementing the backpropagation algorithm so the layers work correctly together, and implementing a number of different neural net training optimization techniques such as Dropout, Momentum, and Weight Regularization. We'll end up with a Jupyter notebook running a flexible deep learning framework live that can then be extended to create arbitrarily deep networks - all from scratch. Attendees will leave this talk with a deeper understanding of how and why neural nets work. This will help them when solving deep learning problems, as well as giving them confidence to explain the inner workings of neural nets to others - whether at conferences, as part of their jobs, or during job interviews.

How Deep Neural Networks Work and How We Put Them to Work at Facebook - Brandon Rohrer, Data Scientist at Facebook

Deep learning is the technology driving today's artificial intelligence boom. It is particularly good at image classification, for instance, deciding whether a picture contains a cat. To get the best results, it's helpful to understand how they work. We will take a gentle, detailed tour though a multilayer fully-connected neural network, backpropagation and a convolutional neural network. You won't need any background in math, programming or machine learning. At Facebook we use deep neural networks as part of our effort to connect the entire world. To provide network connectivity to everyone, we first have to know where everyone is. We use deep learning to find buildings in satellite images so that we know which network technologies will work best and where to deploy them.

Deep Learning in the real world - Lukas Biewald, Founder of CrowdFlower

As company take machine learning out of R&D and into production they face a whole set of new challenges. Human in the loop, active learning and transfer learning are all essential design patterns for making deep learning real.


Deep Learning for Practical Natural Language Processing - Luke de Oliveira, Co-founder, Head of Research at Vai Technologies

Natural language processing (NLP) is one of the most transformative technologies for modern businesses and enterprises. We will focus on practical applications and considerations of applying deep learning for NLP in industrial settings. This hands-on session will target algorithms and frameworks which are able to be deployed in products today. We will provide an overview the different components that go into end-to-end deep learning systems, including word vector representations (word2vec, GloVe, fastText, etc.), recurrent neural networks, convolutional neural networks, and attention mechanisms. We will also cover selected tips and tricks for ensuring deep learning products add maximal value in application architectures, as well as provide some guidelines for managing NLP systems in the wild.

Deep Learning Pipelines on Apache Spark - Joseph Bradley, Apache Spark PMC member and Machine Learning Software Engineer at Databricks

Deep learning has allowed remarkable results in fields such as Computer Vision and Natural Language Processing, but barriers to entry inhibit many data scientists from using deep learning as an everyday tool. Existing deep learning frameworks require significant programming, and scaling up via distributed computing requires even more work. In this talk, we discuss our new open source library meant to address some of these challenges.

Deep Learning Pipelines is an open source library integrating popular deep learning libraries with Apache Spark. The library aims to help engineers and data scientists familiar with Spark to train and deploy deep learning models into their existing workflows. This package simplifies deep learning in three major ways: