Hands-on Tutorials

We have a fantastic lineup of hands-on tutorials to be held in conjunction with KDD 2019. Check back as we get closer to the conference for more detailed program information.

  • Put Deep Learning to Work: A Practical Introduction using Amazon Web Services

    Machine learning (ML) and deep learning (DL) projects are becoming increasingly common at enterprises and startups alike and have been a key innovation engine for Amazon businesses such as Go, Alexa, and Robotics. In this workshop, Wenming Ye (AWS) and Miro Enev (Nvidia) offer a practical next step in DL learning with instructions, demos, and hands-on labs. You will explore the current trends powering AI/DL adoption and algorithmic learning in neural networks, dive into how DL is applied in modern business practices, and leverage building blocks from the Amazon ML family of AI services from powerful new GPU instances, convenient Amazon SageMaker built-in Algorithms, to ready-to-use managed AI services.

    Presenters:

    Wenming Ye (Amazon Web Services) & Miro Enev (Amazon Web Services)


    Timeslot: Tue, August 06, 2019 - 9:30 am - 12:30 pm
    Tutorial Resources

  • Democratizing & Accelerating AI through Automated Machine Learning

    Democratizing & Accelerating AI through Automated Machine Learning. During this hands-on workshop, the attendees will learn how to use Automated ML for model training and Azure ML SDK for deployment of this trained model as a webservice.

    Presenters:

    Parashar Shah (Microsoft) & Krishna Anumalasetty (Microsoft)


    Timeslot: Tue, August 06, 2019 - 9:30 am - 12:30 pm
    Tutorial Resources

  • Concept to Code: Deep Neural Conversational System

    Deep Neural Networks (DNNs) are emerging as the most powerful technique that can achieve excellent performance on difficult learning tasks. In this hands on tutorial we present necessary concepts for understanding and implementing conversational system based on DNNs (or more specifically, Recurrent Neural Network (RNN)). We begin the hands-on tutorial with concepts for core building blocks which include (a) word embeddings like Word2Vec, (b) different RNNs like LSTM, GRU etc. and (c) different attention mechanisms. Subsequently, we cover standard approaches including sequence to sequence (seq2seq) framework and seq2seq with attention mechanisms. Next, we cover dialog generation based on hierarchical neural network, memory network and end-to-end memory network. Finally, we touch upon advanced approaches including deep reinforcement learning for dialog generation, conversational QA over a large scale knowledge base and application of Bidirectional Encoder Representation from Transformer (BERT) for dialogue generation.

    We conduct hands on session for two implementations of conversation system based on neural conversation model & attention and memory networks. We provide details for installation prerequisite and code using Jupyter notebooks with comments on concepts, key steps, visualization and results.

    We believe that a self-contained hands-on tutorial giving good conceptual understanding of core techniques with sufficient mathematical background along with actual code will be of immense help to participants.

    Presenters:

    Omprakash Sonie (Flipkart), Abir Chakraborty (Flipkart), Nikesh Garera (Flipkart)


    Timeslot: Tue, August 06, 2019 - 9:30 am - 12:30 pm
    Tutorial Resources

  • Cloud-Based Data Science at the Speed of Thought Using RAPIDS - the Open GPU Data Science Ecosystem

    The RAPIDS suite of open source software libraries gives the data scientist the freedom to execute end-to-end data science and analytics pipelines on GPUs. RAPIDS is incubatedby NVIDIA based on years of accelerated analytics experience. RAPIDS relies on NVIDIA CUDA primitives for low-level compute optimization and exposes GPU parallelism and high-bandwidth memory speed through user-friendly Python interfaces. Through a familiar DataFrame API that integrates with a variety of machine learning algorithms, RAPIDS facilitates common data preparations tasks while removing typical serialization costs. RAPIDS includes support for multi-GPU deployments, enabling vastly accelerated processing and training on large dataset sizes.

    Join NVIDIA’s engineers as they walk through a collection of data science problems that introduce components and features of RAPIDS, including: feature engineering, data manipulation, statistical tasks, machine learning, and graph analysis. This tutorial focuses on accelerating a large data science workflow in Python on a multiple GPU.

    Presenters:

    Brad Rees, PhD (NVidia), Bartley Richardson, PhD (NVidia), Tom Drabas, PhD (Microsoft), Keith Kraus (NVidia), Corey Nolet (NVidia), & Juan-Arturo Herrera, PhD (Microsoft)


    Timeslot: Tue, August 06, 2019 - 1:30 pm - 4:30 pm
    Tutorial Resources

  • Introduction to computer vision and realtime deep learning-based object detection

    Ever wonder how the Tesla Autopilot works (or why it fails)? In this tutorial we will look under the hood and build some of this tech pipeline in a Jupyter Notebook using Python, OpenCV, Keras and Tensorflow. Computer vision (CV) has been revolutionized by deep learning in the past 7-8 years. Exciting real world deployments of computer vision are appearing in the cloud and on the edge. For example, autonomous vehicles, face detection, checkout-less shopping, security systems, cancer detection and more. In this tutorial, we will briefly overview the basics of computer vision before focussing on object detection, where we present modern day pipelines that are being used in application areas, such as, advanced driver assistance systems (ADAS), driver monitoring systems (DMS), and security and surveillance systems. These pipelines are based on complex deep convolutional neural network (CNN) architectures (often 50-60 layers deep), multi-task loss functions, and are either two-stage (e.g., Faster R-CNN) or single-stage (e.g., YOLO/SSD) in nature. We will demonstrate in a Jupyter notebook how to build, train, and evaluate computer vision applications with a primary focus on building an object detection application from scratch to detect logos in images/video. We will also look at training a more general purpose object detection system from scratch and applying it to images/video in real time.

    Presenters:

    Dr. James G. Shanahan (Church and Duncan Group, University of California, Berkeley) & Liang Dai (Facebook, University of California Santa Cruz)


    Timeslot: Tue, August 06, 2019 - 1:30 pm - 4:30 pm
    Tutorial Resources

  • Declarative Text Understanding with SystemT

    With the proliferation of information in unstructured and semi-structured form, text understanding (TU) is becoming a fundamental building block in enterprise applications. SystemT is an industrial-strength system for developing end-to-end TU applications in a declarative fashion. Commonly used text operations are abstracted as built-in operators with clean, well-specified semantics exposed through a formal declarative language called AQL (Annotation Query Language), which are transformed by a compiler into highly optimized internal implementations and executed by an efficient and lean runtime engine. Its architecture is designed to address the main requirements for enterprise TU: scalability, expressivity, transparency, and extensibility. SystemT today ships with multiple products across 4 IBM Software Brands and is used in multiple ongoing research projects and being taught in universities. Ongoing research and development efforts focus on making SystemT more usable for both technical and business users, and continuing enhancing its core functionalities based on natural language processing, machine learning, and database technology.

    In this tutorial, we will motivate declarative TU and showcase SystemT features for meeting the enterprise TU requirements. The hands-on labs will provide concrete examples on performing TU in progressively sophisticated use cases.

    Presenters:

    Huaiyu Zhu (IBM Research Almadean), Yunyao Li (IBM Research – Almaden), Laura Chiticariu (IBM), Marina Danilevsky (IBM Research – Almaden), Sanjana Sahayaraj (IBM Research – Almaden), & Teruki Tauchi (IBM)


    Timeslot: Tue, August 06, 2019 - 1:30 pm - 4:30 pm
    Tutorial Resources

  • Deep Learning for NLP with TensorFlow

    TensorFlow will be used for a hands-on experience with the latest NLP techniques. Starting with a practical introduction to the framework, attendees will get familiar with core APIs, Colab and best practices for data and training pipeline. The classic NLP topics of Embeddings, seq2seq, attention and Neural Machine Translation will be covered, as well as the modern deep learning architectures of Transformer and BERT. Participants will get exposed to foundational NLP theory and state-of-the-art models, understand them conceptually and apply them to practical problems, for instance by fine-tuning pretrained BERT models.

    Presenters:

    Cesar Ilharco Magalhaes (Google), Gabriel Ilharco Magalhaes (Google), & Jason Baldridge (Google)


    Timeslot: Wed, August 07, 2019 - 9:30 am - 12:30 pm
    Tutorial Resources

  • Deep Learning at Scale on Databricks

    In deep learning, the bigger the data, the better. However, scaling your traditional deep learning tools to meet the ever increasing size of data is difficult. Enabling other data scientists to reproduce your model and effectively train and test your model at scale is much harder. This hands-on tutorial will teach you how to do production level Deep Learning at scale on Databricks. It will guide you through the steps of building a single node deep learning model to distributed model inference and finally, distributed model training and model productionization. This tutorial will provide hands-on experience with:

    • Single-node deep learning concepts (Keras)
    • Tracking experiments and reproducing machine learning models (MLflow)
    • Scalable model inference (Apache Spark and MLflow)
    • Distributed model training (Horovod and Keras)

    By the end of this session, you will be able to build your own deep learning model using Keras, track and reproduce your experiments with MLflow, perform distributed inference using Apache Spark, and build a distributed deep learning model using HorovodRunner. Further, you will be able to identify when/where distributed deep learning training should be applied and common techniques to optimize training of distributed models.

    Presenters:

    Amir Issaei (Databricks) & Brooke Wenig (Databricks)


    Timeslot: Wed, August 07, 2019 - 9:30 am - 12:30 pm and 1:30 pm - 4:30pm
    Tutorial Resources

  • Deep learning for time series forecasting

    Time series forecasting is one of the most important topics in data science. Almost every business needs to predict the future in order to make better decisions and allocate resources more effectively. Examples of time series forecasting use cases are: financial forecasting, product sales forecasting, web traffic forecasting, energy demand forecasting for buildings and data centers and many more. However, most existing forecasting solutions use traditional time series and machine learning models. For complex forecasting problems, data scientists need to know how to leverage advanced techniques to generate more accurate forecasts.

    Deep neural networks have achieved a lot of success for many applications. In particular, recurrent neural networks (RNNs) are frequently used in text, speech and video analysis, being designed for processing sequential data. Additionally, convolutional neural networks (CNNs) have achieved state-of-the-art performance on many computer vision tasks. These methods, as well as hybrid models that combine deep neural networks with traditional statistical algorithms, have only recently been applied to the task of time series forecasting. Nevertheless, they achieved state-of-the-art performance in several benchmarks and won multiple competitions. In this tutorial, we describe the basic concepts for building such models and demonstrate how and when to apply them to time series forecasting.

    Presenters:

    Yijing Chen (Microsoft), Dmitry Pechyoni (Microsoft), Angus Taylor (Microsoft), & Vanja Pauni (Microsoft)


    Timeslot: Wed, August 07, 2019 - 9:30 am - 12:30 pm
    Tutorial Resources

  • Introduction to Kubeflow Pipelines

    We will walk through Kubeflow Pipelines to create a full machine learning application on Kubernetes so you can become familiar with Google Cloud Platform tools such as Cloud Shell and Kubernetes Engine. You’ll start with an empty environment and create a Kubernetes cluster and install Kubeflow from scratch. You can build and run a full pipeline that does distributed training of a TensorFlow model, then scales and serves the trained model and deploys a web frontend for requesting predictions from the model. Finally, we will teach you how to use a Jupyter notebook to build and run a pipeline using the Kubeflow Pipelines SDK.

    Presenters:

    Dan Anghel (Google), Michael Jastrzebski (GitHub), & Hamel Husain (GitHub)


    Timeslot: Wed, August 07, 2019 - 1:30 pm - 4:30 pm
    Tutorial Resources

  • Building production-ready recommendation systems at scale

    Recent decades have witnessed a great proliferation of recommendation systems. The technology has brought significant profits to many business verticals. From earlier algorithms such as similarity based collaborative filtering to the latest deep neural network based methods, recommendation technologies have evolved dramatically, which, to some extent, makes it challenging to practitioners to select and customize the optimal algorithms for a specific business scenario. In addition, operations such as data preprocessing, model evaluation, system operationalization etc. play a significant role in the lifecycle of developing a recommendation system; however, they are often neglected by practitioners.

    Based on extensive experience in productization of recommendation systems in a variety of real-world application domains, in this tutorial, we will review and demonstrate the main key tasks in building recommendation systems. We will use best practices and will provide examples of democratizing recommendation systems to every organization and the wider community. An open source GitHub repository, Microsoft/Recommenders, where the key topics are shared as Jupyter notebooks and a utility function codebase, will be used for the hands-on practice. This repository is designed to help data scientists quickly grasp basic concepts in a hands-on fashion and has gained good visibility within the community with more than 2,500 stars on GitHub. We will walk through several recommendation algorithms in order to provide an in-depth understanding of the techniques. We believe that the best practice examples shared in the repository will help developers / scientists / researchers to quickly build production-ready recommendation systems as well as to prototype novel ideas using the provided utility functions.

    Presenters:

    Le Zhang (Microsoft), Tao Wu (Microsoft), Xing Xie (Microsoft), Andreas Argyriou (Microsoft), Miguel Fierro (Microsoft), & Jianxun Lian (Microsoft Research Asia)


    Timeslot: Thu, August 08, 2019 - 1:00 pm - 4:00 pm
    Tutorial Resources

  • Learning Graph Neural Networks with Deep Graph Library

    Learning from graph data has played a substantial role in many real world scenarios including social network analysis, knowledge graph construction, protein function prediction and so on. Recent burst of researches on Graph Neural Networks (GNNs) brings representation learning to non-euclidean space and achieves state-of-art results in community detection, drug discovery, recommendation, etc. More recent perspective begins to view GNN as a more general form of the neural network models, such as attention architecture, that have dominated areas of computer vision and natural language processing. As graph is essentially relation, modeling explicit or inferring latent graph structure is crucial to the ability of relational reasoning for model AI.

    Presenters:

    Minjie Wang (NYU), Lingfan Yu (NYU), Da Zheng (AWS), & Nick Choma (NYU)


    Timeslot: Wed, August 07, 2019 - 1:30 pm - 4:30 pm
    Tutorial Resources

  • From Shallow to Deep Language Representations: Pre-training, Fine-tuning, and Beyond

    Natural language processing (NLP) is at the core of the pursuit for artificial intelligence, with deep learning as the main powerhouse of recent advances. Most NLP problems remain unsolved. The compositional nature of language enables us to express complex ideas, but at the same time making it intractable to spoon-feed enough labels to the data-hungry algorithms for all situations. Recent progress on unsupervised language representation techniques brings new hope. In this hands-on tutorial, we walk through these techniques and see how NLP learning can be drastically improved based on pre-training and fine-tuning language representations on unlabelled text. Specifically, we consider shallow representations in word embeddings such as word2vec, fastText, and GloVe, and deep representations with attention mechanisms such as BERT. We demonstrate detailed procedures and best practices on how to pre-train such models and fine-tune them in downstream NLP tasks as diverse as finding synonyms and analogies, sentiment analysis, question answering, and machine translation. All the hands-on implementations are with Apache (incubating) MXNet and GluonNLP, and part of the implementations are available on Dive into Deep Learning (www.d2l.ai).

    Presenters:

    Zhang, Aston (Amazon AI), Zha, Sheng (Amazon AI), Lin, Haibin (Amazon AI), Smola, Alexander (Amazon AI), Li, Mu (Amazon AI), & Wang, Chenguang (Amazon AI)


    Timeslot: Thu, August 08, 2019 - 9:30am - 12:30 pm | 1:00 pm - 4:00 pm
    Tutorial Resources

  • From Graph to Knowledge Graph: Mining Large-scale Heterogeneous Networks Using Spark

    Many real-world datasets come in the form of graphs. These datasets include social networks, biological networks, knowledge graphs, the World Wide Web, and many more. Having a comprehensive understanding of these networks is essential to truly understand many important applications.

    This hands-on tutorial introduces the fundamental concepts and tools used in modeling large-scale graphs and knowledge graphs. The audience will learn a spectrum of techniques used to build applications that use graphs and knowledge graphs: ranging from traditional data analysis and mining methods to the emerging deep learning and embedding approaches.

    Five lab sessions are included to give the audience hands-on experience to work through real-life examples on major topics covered in this tutorial. This includes: (1) understanding basic graph properties; (2) using graph representation learning to explore network similarity; (3) utilizing NLP and text mining techniques to build knowledge graphs; (4) modeling knowledge graphs with embedding techniques and how to apply it to recommendation applications.

    We use Microsoft Academic Graph (MAG)—the largest publicly available academic domain knowledge graph –- as the dataset to demonstrate the algorithms and applications presented here. MAG includes 6 types of entities with 450 million nodes, and over 3 billion edges covering more than 660K academic concepts. The MAG dataset (500G+) is regularly updated at a bi-weekly cadence. We use a Top CS Conference Sub-Graph from one of the most up-to-date data versions for this hands-on tutorial.

    Presenters:

    Iris Shen (Microsoft Research), Charles Huang (Microsoft Research), Chieh-Han Wu (Microsoft Research), Anshul Kanakia (Microsoft Research) 


    Timeslot: Thu, August 08, 2019 - 9:30 am - 12:30 pm | 1:00 pm - 4:00 pm
    Tutorial Resources

  • Practice of Efficient Data Collection via Crowdsourcing at Large-Scale

    In this tutorial, we present you a portion of unique industrial practical experience on efficient data labeling via crowdsourcing shared by both leading researchers and engineers from Yandex. Majority of ML projects require training data, and often this data can only be obtained by human labelling. Moreover, the more applications of AI appear, the more nontrivial tasks for collecting human labelled data arise. Production of such data in a large-scale requires construction of a technological pipeline, what includes solving issues related to quality control and smart distribution of tasks between workers.

    We will make an introduction to data labeling via public crowdsourcing marketplaces and will present key components of efficient label collection. This will be followed by a practical session, where participants will choose one of real label collection tasks, experiment with selecting settings for the labelling process, and launch their label collection project at Yandex.Toloka, one of the largest crowdsourcing marketplace. The projects will be run on real crowds within the tutorial session. Finally, participants will receive a feedback about their projects and practical advices to make them more efficient. We invite beginners, advanced specialists, and researchers to learn how to collect labelled data with good quality and do it efficiently.

    Presenters:

    Alexey Drutsa (Yandex), Valentina Fedorova (Yandex), Evfrosiniya Zerminova (Yandex), & Olga Megorskaya (Yandex)


    Timeslot: Thu, August 08, 2019 - 9:30 am-12:30 pm
    Tutorial Resources

Diamond Sponsors

Platinum Sponsors

Gold Sponsors

Silver Sponsors

Bronze Sponsors

Sponsor KDD 2019

How can we assist you?

We'll be updating the website as information becomes available. If you have a question that requires immediate attention, please feel free to contact us. Thank you!

Please enter the word you see in the image below: