The problem with all learning is that only practice can validate whether you actually learned something. Let’s see how could you learn ML and DL without having to install anything – just in cloud and for free!
This blog item is not about how to learn Machine Learning or Deep Learning (which is subset of Machine Learning but recently so big that constitutes its own area). For that see for example this article from Forbes or bing it. This blog item is about 3 things only: practice, practice, practice.
So: this blog item is about learning Machine Learning and Deep Learning in practice. Of course you can take one of many available courses online and you can read one of many books from Amazon Kindle, but at the end of the day if your knowledge is theoretical it’s like it doesn’t even exist.
In other words: you can say that you know Machine Learning (ML) and Deep Learning (DL) only if you can apply your knowledge in practice what in the world of ML+DL usually means:
- You can prepare datasets for training, for example by data cleaning or feature engineering
- You know how to train and tweak learning algorithms with use of ML/DL libraries in programming languages Python or R which are de-facto standard for data scientists (a name commonly used nowadays for expert in ML/DL, although some Data Scientists know only about Big Data but not so much about ML/DL)
- You know how to execute predictions/inferences based on trained models
Basically in the past the only choice to practice ML was to install environment with Python and R and related libraries and tools locally in your Windows PC or macOS macbook and then play around in it. However nowadays mobility – the fact that people prefer to work on tablets like iPad and on smartphones like Android – means that it would be pretty ideal to practice machine learning also in the cloud, and using mobile devices merely as user interface for example with help of mobile web browser.
Most popular tool for Data Scientists practicing Machine Learning and Deep Learning is Jupyter (like Jupiter but with “y” in it because data scientists love “Python” programming language). Jupyter is kind of a live/dynamic (not static!) notebook that runs as local web server in your PC or Mac and you use it via web browser on local host by entering lines of Python code – line by line.
While Kaggle – very popular website where people can participate in competitions in Machine Learning – has so called Kaggle Kernels, that are basically a web-based Jupyter notebooks, recently also Microsoft started offering free cloud-based Jupyter notebooks with support for Python, R, SciKit-Learn (popular library for classic Machine Learning), TensorFlow (popular library for Deep Learning): Azure Notebooks.
Since nowadays in practice a lot of training of Deep Learning models takes place in the cloud anyway, while models are used for prediction/inference in edge devices like smartphones, having ability to learn ML/DL in the cloud prepares you also for the future as training of models in the cloud instead on premises will be increasingly important. (Although note that Deep Learning ability in edge devices increases also as both Apple and Huawei already are selling smartphones with built-in hardware-based Deep Learning)
So how are Azure Notebooks working? You gotta go to https://notebooks.azure.com and then log-in with Microsoft account and then create library and there you can create Jupyter notebook. When you open Jupyter notebook on Azure Notebooks the server is being started:
… and then soon you can start playing around (note which versions of SkiKit-Learn and TensorFlow are offered now):
Conclusion: running Jupyter on the cloud let’s you learn Machine Learning and Deep Learning without the need to install anything on your PC or Mac – so for example also via iPad or Chromebook!