Thursday 30 March 2023

Machine Learning on Cloud


What is machine learning?

Humans learn by experience. Humans have the capability of making decisions based on their past experiences. But what about machines?? Can they think and make decisions like humans do? Well there comes into picture MACHINE LEARNING! Machine learning is a field where machine learns using historical data, understands the data pattern and using this as experience gains capability to make future decisions without being explicitly programmed.  

Suppose you are using an OTT platform, say Netflix. The first time you login and search for a movie or particular genre,actor or actresses, search results are displayed on the screen. But the next time you revisit your account, the home screen displays all the content related to your previous searches. This is done using ML. Machine learns from your search history and your watching pattern and suggests related content to you.

Machine learning helps in sales forecasts helping them plan their further strategies and make production plans. Also recommends the right products to the customers, simplify marketing, help plan offers ,etc  leading to more profit. Forecasting weather can help authorities to make disaster management beforehand in case of any natural disasters forthcoming. It can be lifesaving when ML is used to predict chances of occurrences of heart diseases, cancers and other fatal diseases.

However, machine learning requires massive infrastructure, programmers familiar with ML, speed, high-end computing, storage, fast performance and data analysis, and large amounts of data to provide these machine learning algorithms, and a local infrastructure can do that. The cloud is The reason why computation is so important in machine learning! The cloud offers the speed and performance of GPUs and FPGAs without having to configure your own infrastructure.Cloud-based machine learning and deep learning platforms tend to provide their own pre-packaged algorithms and models, and support some external frameworks or containers with specific entry points to make work easier . They also provide AI services optimized for use cases such as computer vision, natural language processing, text-to-speech, and predictive analytics.


GPUaaS, GPU as a Service, lets you focus on building, training, and deploying AI solutions for end users. This eliminates the need to configure GPU infrastructure locally. Compute-intensive tasks consume significant CPU resources. GPUaaS helps you free up resources and increase performance by offloading some of the work to the GPU. This helps reduce costs associated with proprietary GPU infrastructure, increases scalability and flexibility, and enables customers to implement large-scale, large-scale GPU computing solutions.

What are cloud computing platforms for machine learning?

There are many cloud computing platforms for machine learning, like AWS, Microsoft Azure, Google Cloud, IBM, etc.

Let's take a quick look at what AWS offers for Machine Learning -

  • An ML-based service, Amazon Forecast uses time-series historical data to predict/predict future outcomes.

  • Amazon Sagemaker is used to build, test, and deploy machine learning models in the cloud.

  • Amazon Translate is a very accurate and fast translation service, allowing you to globalize your content. Supports 5,500 translation combinations and accurately identifies the source language.

  • Amazon Personalize is a recommendation service that uses our data to provide recommendations. It revolves around the individual.

  • Amazon Polly is a machine learning service that turns speech into realistic speech. Polly currently has 47 male and female voices and speaks 24 languages.

Workflow Services

  • AWS Deep Learning AMI

  • AWS DL Containers

  • AWS Batch

  • Parallel Clusters are workflow services available on Amazon.


The various frameworks supported byAWS are -

  • TensorFlow is an open-source, Python-compatible library for numerical computation that makes machine learning and neural network development faster and easier.

  • PyTorch on AWS is an open source deep learning framework that makes it easy to develop machine learning models and deploy them in production.

  • MXNet is an open-source deep learning framework that lets you define, train, and deploy deep neural networks on devices ranging from cloud infrastructure to mobile devices.

  • Keras is an open source software library that provides a Python interface to artificial neural networks.

  • Gluon is a new deep learning library that enables developers to prototype, build, train, and deploy machine learning models for cloud, edge, and mobile applications.


EC2 p4, EC2 p3, EC2 g4, EC2 inf1 are instances  specially used for ml and ai.


  • Amazon Elastic Inference lets you pair low-cost GPU-powered acceleration with Amazon EC2 and Sagemaker instances or Amazon ECS workloads to reduce the cost of running inference by up to 75% deep learning.

  • Elastic Fabric Adapter (EFA) is a network interface for Amazon EC2 instances that allows customers to run applications that require high levels of inter-node communication at scale on AWS.


Large amounts of data used in ML and AL can be stored in AWS S3, AWS EBS, AWS FSx, AWS EFS, etc.

This article discusses several machine learning workload offers from AWS. There are numerous other services that support ML. So, it makes sense to run ML workloads in the cloud given its strong performance, compute, speed, and storage capabilities.

Written by Rajeshwari Jedhe, Research Analyst - Contact Discovery (


No comments:

Post a Comment

Maximizing Content Delivery Efficiency: Optimizing Performance in AWS CloudFront

  Unleash Blazing-Fast Content Delivery: Your Guide to CloudFront Optimization Introduction: AWS CloudFront stands as a cornerstone of moder...