A Platform for Machine Learning
linkedin
December 03, 2018
Our general assumption is that Data Science is all about Algorithm development and Model selection. Not quite true in today's BigData world. If you have a neatly packaged CSV file as data source then yes, you can spend quality time figuring out how different features correlate with each other and extract patterns using Machine Learning (ML). However, today’s data Scientist usually spends more than 50% time dealing with other Concerns (areas of focus) like Data Preparation, Cleansing, Distributed Training, Model Deployment and Scaling.
See publication
Tags: AI, Big Data, Cloud
To GPU or TPU...that's the question!
linkedin
October 26, 2018
DISCLAIMER: This post is not a reflection of views of General Electric. This is just a personal hobby post I have written to share knowledge on this topic.
It's pretty much universally accepted that to effectively Train large Deep Learning Models we need some sort of hardware acceleration over the CPU. Models with many Neural Layers of Learning like Convolutional Neural Nets (best for image analysis) and Recurrent Neural Nets (for sequence data like text or time-series) - involve huge number of parallel operations during Training and Inference. The hardware acceleration engines have 1000s of cores for handling basic Linear Algebra calculations in parallel. NVIDIA is the leader in this space with their tried and tested GPUs. GPUs come in many form factors to server processing needs in the Data Center and for Edge devices.
See publication
Tags: Analytics, AR/VR, Big Data
Quick (and free) experiment for CPU vs GPU for Deep Learning
linkedin
August 01, 2018
Recently I presented at some NVIDIA conferences and got many questions on GPU vs CPU - Is the GPU really worth it? Does it really provide the benefits advertised? To be honest, we were skeptical before getting a desktop GPU for training our Video Analytics models. However, once we started using it and saw ours models zip through the training process in seconds compared to minutes (and sometimes hours) - we were definitely convinced. You need to have a solid case where you will be continuously training models to justify a local version - but for general experimentation you can always start with a Cloud instance.
See publication
Tags: Analytics, AI, Cloud
Statistics, Machine Learning and a scenic Train journey
linkedin
June 14, 2018
I have been preparing content for an Analytics class I will be teaching shortly. During this curation process had some interesting discussions with folks on real-world use-cases and all the jargon like Null hypothesis, Hypothesis testing, p-Value, Z-score, Supervised, Unsupervised, Training & Test datasets, Correlations, etc. Figured I should consolidate my learnings into this post and get views from a wider audience. Mainly to get your feedback and opinions and see if I got anything wrong (very possible).
See publication
Tags: Analytics, AR/VR
Listen to your machines using Predix
linkedin
January 31, 2018
Check out my latest Blog post on Predix.io. Was a simple experiment we started with on collecting and analyzing machine sounds in Predix Cloud. Many more things can be done with this audio data - Lots of potential here!
See publication
Tags: AI, Cloud
Deep Learning on Predix for Computer Vision
linkedin
November 06, 2017
Hi folks - I just published a video tutorial showing a very simple way to build a web application on GE Predix Platform. This app reads images uploaded and extracts Knowledge from these using a pre-trained Deep Learning model. This application can detect major objects in the image and highlight them with a degree of confidence.
See publication
Tags: Analytics, AI