Microsoft Azure Machine Learning

What differentiates Microsoft Azure Machine Learning from other similar products?

training neural networks is very resource-intensive with hardware, and this platform makes it so simple to access the hardware you need. and it is the best part since it is easy to run distributed training on numerous nodes without ever having to write that complex network code.

What is your favorite aspect of this product?

i personally enjoy the hyperdrive feature of automated parameter tuning since it explores the search space in a significantly more efficient way than i myself could. it also saves a significant amount of money in that it can pause low performing runs early to save compute credits and really makes it easy to deploy our models in kubernetes. it feels like a robust and professional system to any team doing high end work in the field of AI research and development.

What do you dislike most about this product?

at times, it feels cluttered through the portal interface, as the interface offers many tabs and options to choose between, so it can be challenging to keep track of the location of a particular setting.

What recommendations would you give to someone considering this product?

when you are lifting big in deep learning this is the tool to work with. it offers the raw power and the managerial capabilities to achieve results.

What differentiates Microsoft Azure Machine Learning from other similar products?

I would spend so much time initially establishing my local environment with the correct versions of libraries but azure ml does all of that in the cloud. It is superior to the standard jupyter since it is attached to all of my data sources in the cloud and accessible compute power. It is the best because i can spin up a massive gpu instance with a single button press whenever i need to train a deep learning model.

What is your favorite aspect of this product?

I like how it manages to store all my history of experiments to make cross-comparing of the variation performance of different code versions simple. The terminal access is also convenient and useful as well when i need to install some custom packages not available in the default environment. The intellisense feature in the notebook editor is also quite surprisingly good and allow me to write faster with few mistakes. It is a well developed development environment.

What do you dislike most about this product?

It can be time consuming to compile the image of the environment. When your dependencies change on a regular basis you will be waiting a long time before the new image is created.

What recommendations would you give to someone considering this product?

This is a phenomenal upgrade to any person living in notebooks as it provides all the cloud power without installing it locally.

What differentiates Microsoft Azure Machine Learning from other similar products?

I do not (like many others do) love code writing all the little stuff and the designer that comes with this tool is precisely what i was in need of. It is superior to the rest in the way that you can easily connect and understand the module, i can see where the data is getting mixed up in the graphical representation.

What is your favorite aspect of this product?

The most satisfying aspect is the library of ready made modules to common operations such as principal component analysis or sentiment detectors. It just feels more akin to a visual orginator with the flexibility of deep customization. It has made my daily workflow more enjoyable and less frustrating.

What do you dislike most about this product?

Occasionally the error messages in the visual designer are a cube. One has to do some searching on what exactly module has failed in the pipeline.

What recommendations would you give to someone considering this product?

In case you need a visual approach to data science this is the most powerful lightweight on the market. It is not complex but simple to learn quickly.