[TALK]@Telematika

dataset

Bootstrap custom image classification using transfer learning

January 2, 2020

Transfer learning enables you to use pretrained deep neural networks (e.g., AlexNet, ResNet, and Inception V3) and adapt them for custom image classification tasks. Danielle Dean and Wee Hyong Tok walk you through the basics of transfer learning and demonstrate how you can use the technique to bootstrap the building of custom image classifiers.

Interactive visualization for data science

December 31, 2019

One of the challenges in traditional data visualization is that they are static and have bounds on limited physical/pixel space. Interactive visualizations allows us to move beyond this limitation by adding layers of interactions. Bargava Subramanian and Amit Kapoor teach the art and science of creating interactive data visualizations.

Managing machine learning models in production

December 30, 2019

There are many challenges to deploying machine models in production, including managing multiple versions of models, maintaining staging and production models, keeping track of model performance, logging, and scaling. Anand Chitipothu explores the tools, techniques, and system architecture of a cloud platform built to solve these challenges and the new opportunities it opens up.

Real-world patterns for continuously deployed advanced analytics

December 30, 2019

How can we drive more data pipelines, advanced analytics, and machine learning models into production? How can we do this both faster and more reliably? Graham Gear draws on real-world processes and systems to explain how it's possible to apply continuous delivery techniques to advanced analytics, realizing business value earlier and more safely.

Cloud architectures for data science

December 27, 2019

Data is available from an incredible number of sources in an endless number of formats. Data science deals with the extraction of valuable insights from this jumble in the form of attractive visualizations. Walking you through several examples using practical tools and tricks, Margriet Groenendijk presents a typical workflow that offers a basic introduction to data science.

Open funding data: 360Giving's journey so far

December 17, 2019

At present, it's not possible to find a complete dataset on all charitable grants in the UK. Alice Casey and Edafe Onerhime offer an overview of 360Giving, which supports funding organizations in publishing their grants data in an open, standardized way and helps people to understand and use that data to support decision making and learning across the charitable giving sector.

Big data meets the IoT

December 15, 2019

The IoT is fundamentally transforming industries and reconfiguring the technology landscape, but challenges exist for enterprises to effectively realize the value from this next wave of information and opportunity. Cheryl Wiebe explores how leading companies harness the IoT by putting IoT data in context, fostering collaboration between IT and OT and enabling a new breed of scalable analytics.

Collaboration and openness drive innovation in artificial intelligence

December 14, 2019

The power of artificial intelligence and advanced analytics emerges from the ability to analyze and compute large, disparate datasets from varied devices and locations, such as predictive medicine and automated cars, at lightning-fast speed. Martin Hall explains why collaboration and openness are the key elements driving innovation in AI.

Creating and evaluating a distance measure

December 14, 2019

Whether we're talking about spam emails, merging records, or investigating clusters, there are many times when having a measure of how alike things are makes them easier to work with (e.g., with unstructured data that isn't incorporated into your data models). Melissa Santos offers a practical approach to creating a distance metric and validating with business owners that it provides value.

How the largest US healthcare dataset in Hadoop enables patient-level analytics in near real time

December 12, 2019

The need to find efficiencies in healthcare is becoming paramount as our society and the global population continue to grow and live longer. Navdeep Alam shares his experience and reviews current and emerging technologies in the marketplace that handle working with unbounded, de-identified patient datasets in the billions of rows in an efficient and scalable way.

The future of column-oriented data processing with Arrow and Parquet

December 9, 2019

In pursuit of speed, big data is evolving toward columnar execution. The solid foundation laid by Arrow and Parquet for a shared columnar representation across the ecosystem promises a great future. Julien Le Dem and Jacques Nadeau discuss the future of columnar and the hardware trends it takes advantage of, like RDMA, SSDs, and nonvolatile memory.

Deep learning at scale and use cases

December 6, 2019

Deep learning has made a major impact in the last three years. Imperfect interactions with machines, such as speech or image processing, have been made robust by deep learning that finds usable structure in large datasets. Naveen Rao outlines deep learning challenges and explores how changes to the organization of computation and communication can lead to advances in capabilities.

Deep neural network model compression and an efficient inference engine

December 6, 2019

Neural networks are both computationally and memory intensive, making them difficult to deploy on embedded systems with limited hardware resources. Song Han explains how deep compression addresses this limitation by reducing the storage requirement of neural networks without affecting their accuracy and proposes an energy-efficient inference engine (EIE) that works with this model.

Deeply active learning: Approximating human learning with smaller datasets combined with human assistance

December 6, 2019

Natural-language assistants are the emergent killer app for AI. Getting from here to there with deep learning, however, can require enormous datasets. Christopher Nguyen and Binh Han explain how to shorten the time to effectiveness and the amount of training data that's required to achieve a given level of performance using human-in-the-loop active learning.

Analytics innovation in cancer research

November 21, 2019

Federated analytics, a new approach to analyzing big data, supports unprecedented collaboration across large distributed datasets that contain proprietary and/or protected information. Gilad Olswang explains how Intel harnesses the power of federated analytics in the Collaborative Cancer Cloud project.

Data relativism and the rise of context services

November 19, 2019

The traditional data warehouse of the 1990s was quaintly called the single source of truth. Joe Hellerstein explains why today we take a far more relativistic view: the meaning of data depends on the context in which it is used.

HopsWorks: Multitenant Hadoop as a service

November 18, 2019

Currently, multitenancy in Hadoop is limited to organizations running separate Hadoop clusters, and the secure sharing of resources is achieved using virtualization or containers. Jim Dowling describes how HopsWorks enables organizations to securely share a single Hadoop cluster using projects and a new metadata layer that enables protection domains while still allowing data sharing.

Introduction to generalized low-rank models and missing values

November 18, 2019

The generalized low-rank model is a new machine-learning approach for reconstructing missing values and identifying important features in heterogeneous data. Through a series of examples, Jo-fai Chow demonstrates how to fit low-rank models in a parallelized framework and how to use these models to make better predictions.

Petascale genomics

November 17, 2019

The advent of next-generation DNA sequencing technologies is revolutionizing life sciences research by routinely generating extremely large datasets. Tom White explains how big data tools developed to handle large-scale Internet data (like Hadoop) help scientists effectively manage this new scale of data and also enable addressing a host of questions that were previously out of reach.

Protecting individual privacy in a data-driven world

November 17, 2019

With the analytic and predictive power of big data comes the responsibility to respect and protect individual privacy. As citizens, we should hold organizations to account; as data practitioners, we must find intelligent ways to analyze data without violating privacy. Jason McFall discusses privacy risks and surveys leading privacy-preserving analysis techniques.

Sightseeing, venues, and friends: Predictive analytics with Spark ML and Cassandra

November 17, 2019

Which venues have similar visiting patterns? How can we detect when a user is on vacation? Can we predict which venues will be favorited by users by examining their friends' preferences? Natalino Busa explains how these predictive analytics tasks can be accomplished by using Spark SQL, Spark ML, and just a few lines of Scala code.

What does your smart car know about you?

November 16, 2019

In the last few years, auto makers and others have introduced devices to connect cars to the Internet and gather data about the vehicles activity, and auto insurers and local governments are just starting to require these devices. Charles Givre gives an overview of the security risks as well as the potential privacy invasions associated with this unique type of data collection.

A few things engineers can learn from designers

October 29, 2019

Big data is great for feeding ML algorithms, but you quickly face a bandwidth issue when interfacing with humans. The brain is a fantastic information-processing machine and has an unparalleled, innate ability to detect patterns. Sbastien Pierre explains what designers can teach engineers about creating new ways to make large volumes of data understandable at the human level.

Amazon for information: Building a modern data catalog

October 28, 2019

A data catalog provides context to help data analysts, data scientists, and other data consumers (including those with little technical background) find a relevant dataset, determine if it can be trusted, understand what it means, and utilize it to make better products and better decisions. Aaron Kalb explores how enterprises build interfaces that make sourcing data as easy as shopping on Amazon.

Atom smashing using machine learning at CERN

October 28, 2019

Siddha Ganju explains how CERN uses machine-learning models to predict which datasets will become popular over time. This helps to replicate the datasets that are most heavily accessed, which improves the efficiency of physics analysis in CMS. Analyzing this data leads to useful information about the physical processes.

Fast data made easy with Apache Kafka and Apache Kudu (incubating)

October 25, 2019

Ted Malaska and Jeff Holoman explain how to go from zero to full-on time series and mutable-profile systems in 40 minutes. Ted and Jeff cover code examples of ingestion from Kafka and Spark Streaming and access through SQL, Spark, and Spark SQL to explore the underlying theories and design patterns that will be common for most solutions with Kudu.

How to make analytic operations look more like DevOps: Lessons learned moving machine-learning algorithms to production environments

October 24, 2019

There is a big difference between running a machine-learning algorithm manually from time to time and building a production system that runs thousands of machine-learning algorithms each day on petabytes of data, while also dealing with all the edge cases that arise. Robert Grossman discusses some of the lessons learned when building such a system and explores the tools that made the job easier.

IoT in the enterprise: A look at Intel (IoT) Inside

October 23, 2019

Moty Fania shares Intels IT experience implementing an on-premises big data IoT platform for internal use cases. This unique platform was built on top of several open source technologies and enables highly scalable stream analytics with a stack of algorithms such as multisensor change detection, anomaly detection, and more.