Wholefoods was where I regularly grocery shopped while in the US – for everything from organic milk, produce, pasta, chocolate to Burt bee’s lip balm. Although pricey, the quality was worth it.
If you didn’t know it already, Amazon, in it’s quest for world dominance bought Whole Foods recently and I’m uncertain how I feel about that. For one they have dropped the prices on some products and giving attractive perks – which is a good thing, but what are we giving up in exchange ? What is Amazon’s long term plan for it ? I certainly hope that Whole Foods continues with the original ethos and encourages organic, sustainable farming practices around the world through it’s suppliers. It will be interesting to see how it all unfolds.
Amazon Machine Learning Conference
Whole Foods carried the most varied selection of fresh herbs. Which I purchased regularly – mostly basil and rosemary, but a few times I’ve picked up organic Sage. So imagine picking up some Sage there, then heading home. Once home, you ask Alexa to help you with a recipe that uses Sage. It could find you a new recipe or it could pull up one that you have already tried and further adapted to your preferences which it has learned over time. Let’s assume you actually finished cooking using the said recipe (Unlike moi who ends up down the rabbit hole of one picture after another of beautifully plated food, from other people’s perfect kitchens and then short on time, fixes something quick. It looks nothing like those pictures, but of course those have been bookmarked for another day…).
After you have finished cooking, you can take pictures or a video with the DeepLens camera which can perhaps tell you the calories and nutrients in the food you just prepared. It can upload those pictures to the AWS cloud, use Amazon SageMaker and run image processing algorithms, video analytics and enhance your picture to look the best or completely transform it into another style through Neural Style Transfer. You could even have an image recognition algorithm to detect the presence of Sage – ‘Sage or no Sage’ similar to ‘Hotdog or not Hotdog’. No I’m not kidding, there’s really an App for that.
So you have gone from holding a bunch of sage leaves in your hand to training models on SageMaker, all without ever leaving the Amazon ecosystem. What’s next? The possibilities are only limited by your imagination. Or so says Amazon.
I realized the ubiquitousness of Amazon yesterday when I had the chance to attend the AWS Innovate 2018 Online conference, Machine Learning special edition. It was a session to create awareness about it’s Machine learning ecosystem and let me tell you – they have you covered.
SageMaker is their flagship offering in the ML suite and was released end of last year. It essentially offers Machine Learning as a service (MLaas). In Amazon’s own words :
Amazon SageMaker is a fully managed end-to-end machine learning service that enables data scientists, developers, and machine learning experts to quickly build, train, and host machine learning models at scale.
As per the keynote, Amazon really wants every developer to be able to use machine learning and they want to be the ones to enable that through their numerous products. They are aiming to remove all the hard edges and obstacles on the path to Machine learning nirvana. In tune with that aim, every step of the SageMaker workflow is optimized to help one get started easily and finish quickly – from data processing, modeling, training to finally deploying the trained models with readily available components, all with just a few clicks. Below is the workflow :
The platform comes with prebuilt notebooks for well known problems that Machine learning tackles. It also comes with about 12 machine learning algorithms prebuilt, both supervised and unsupervised. It also allows one to use custom algorithms and frameworks using a containerized approach with Docker containers. The training can be launched with a single click. One of the neat things is that it offers automated ‘Hyperparameter Optimization (HPO)‘ which should result in saving a lot of time for developers. Once done training, deployment is done with a single click and the hosting is managed by AWS with automatic scaling as required.
The SageMaker is free to try for the first two months right now. Below are the details :
250 hours per month of t2.medium notebook usage for the first two months
50 hours per month of m4.xlarge for training for the first two months
125 hours per month of m4.xlarge for hosting for the first two months
I am certainly excited about what we can do with it and definitely will be making use of this offer. In addition to SageMaker, they covered a whole suite of other interesting and cool products. I hope to cover them in future blog posts.
Go forth and Machine Learn, courtesy the wise Amazon SageMaker.