Convolutional Neural Networks and Using C++ in R with Rcpp

September 24, 2019

Details

Introduction

Ash will cover the architecture of Convolutional Neural Networks (CNN). The Convolution and Pooling operations will be covered. Tools for building a CNN are R and TensorFlow. The implementation of CNN in R will be demonstrated.

Luke will talk about using C++ in R. Are is a beautiful language but some operations can be slow. The RCpp package makes this an easy talk and can drastically improve the performance of your code.

Come early, network, enjoy the food and talks.

Schedule

6:30 - 6:50 Networking 6:50 - 7:00 Welcome & general announcement 7:00 - 7:30 Convolutional Neural Networks 7:30 - 8:00 Speeding up R with Rcpp 8:00 - Raffle 8:00 - 8:30 Networking and Clean-up

Talk 1

Title: Convolutional Neural Networks

Speaker: Ash Pahwa

Abstract

A convolutional neural network (CNN) is a class of deep neural networks used primarily for analyzing visual imagery. CNNs are used for object recognition in an image. Autonomous cars use CNN technology to navigate smoothly on the roads. There are many biomedical applications where CNN is used for image analysis.

The motivation for CNN originated from the Hubel & Wiesel (Noble prize winners 1981) study of the human visual system. This research leads to the idea that if we apply the convolution operations to the image and then feed them to neural networks, the accuracy of object identification increases. This idea started the revolution of CNN.

This talk will cover the architecture of CNN. The Convolution and Pooling operations will be covered. Tools for building a CNN are R and TensorFlow. The implementation of CNN in R will be demonstrated.

Talk 2

Title: Speeding up R with Rcpp

Speaker: Luke Klein

Abstract

R is a beautiful and flexible language, and it is improving all the time. However, some processes (e.g. repeated function calls) can add a lot of computational overhead. Rcpp is a package for extending R with C++ functions which is utilized by over 1000 packages on CRAN. Many R data types and objects can be mapped back and forth to C++ equivalents which facilitates both writing new code for people just learning C++, as well as easier integration of third-party libraries. In this talk, I’ll demonstrate how easy it is to get started with Rcpp and what kind of performance boosts one can achieve. If you’re taking aim at levelling-up your function writing skills then Rcpp is the place to start.

Sponsor

Kforce is an international staffing firm specialized in the IT and Finance and Accounting space. We work with 70% of fortune 100s as well mid to smaller businesses in local markets. We build relationships and are genuinely interested in what you do. If you are looking for a new opportunity or need assistance in finding top talent we can align you to that goal.

Download our Job Board app and receive referral fees on average $500-$1000 https://kfrc.co/download

Contact: John Nacy – Managing Director of Technology at Kforce, [masked]

comments powered by Disqus