Here I have structured my articles according to the different categories they belong to. They do not represent an exhaustive Machine Learning curriculum but they are topics of interest I came across through my Data Science journey.
Python and Machine Learning Initiation
This article is a good start for anyone who would like to give a try at using Python to run a first and simple Machine Learning model.
Data Preparation
This article proposes a mental representation of n-dimensions arrays (also called tensors), which can be useful when dealing with complex datasets for deep-learning projects.
This article is an overview of the advantages of dimensions’ reduction and what they bring when performed in Manufacturing. …
What 23.000 visits in 6 months taught me about the best days to post articles and the level of audience you need to make 10, 100, or 1000$!
I guess that most of the Medium writers followed the same journey as mine: you discover Medium, it does not take long before you get a membership as the content brought by the community is really valuable and, one day, you decide to bring your own piece to the publications you love.
When you are ready to submit your first article, Medium offers you to make your work eligible to the “Medium Partner Program” which lets you place the content behind a paywall and earn a small share of the members’ subscription. Sounds great! …
Hands-on Tutorials
You have probably heard about LIME, SHAP, or Grad-CAM libraries and how they can help you spot areas of interest used by deep learning models for computer vision or object detection… but did you ever thought about making your own? Here is a step-by-step tutorial!
Image by Author
A few weeks ago, I started working on a new project: using cameras and deep-learning to evaluate the level of product in a silo. By monitoring this level, we could run specific operations only when needed, instead of a predefined frequency:
Feel the thrill and build your first A.I. model using Python even if you have no coding experience… yet!
Lately, I had to design a simple coding exercise to help pure beginners take their first steps on Python and feel the “Machine Learning fever”. Today, I am happy to share it with anyone interested to start exploring this field.
And if you ever find yourself stuck with the code, the complete notebook is stored here.
Coding environment
Like for every journey, you will need a proper vehicle to navigate towards your first model.
My advice would be to start with an on-line Python environment like Google Colab.
This will save you from installing Python on your computer or managing libraries (deciding to switch to Pandas 1.0 or the last Scikit-Learn release is probably not your main concern at this time !). …
Sure, it is easy to see how a scalar, a vector, or a matrix look like! But when it comes to n-dimensional arrays, our “narrow” three-dimensional mind can get stuck! But there are ways to go beyond that limitation… and cities are one of them!
The importance of Arrays in Data Science
All Data Science projects rely on structured data (at least at some point!) and some of the usual libraries to handle this information are called “Numpy” or “Pandas”.
The first one (NumPy) is more generic as it allows to deal with multi-dimensional datasets whereas Pandas offers a nice way to handle tabular (2 dimensions) datasets with indexes and specific functionalities.
Like every year, the night of the 24th of December is going to be crazy for Santa, Rudolf, and the elves… But did you know that they optimize their journey around the globe with Google dedicated Operational Research tools?
No matter whether you have been nice this year, following PEP8 rules, or naughty, not even commenting your code… Santa is coming to town!
And when it comes to environmental questions, Santa also aims at being carbon neutral!
A few decades ago, there was still a whole Department at the North Pole taking care of preparing each Christmas journey but, like for every aspect of our society, Data Science has entered into the game! …
Where I share what 20 years of managing tons of messages every day taught me and how to get back in control when you have 103,581 unread emails and counting!
Yes, I know, “20 years of email management” sounds like a veteran’s speech but I started crawling the Web with Netscape in the late ’90s, running a 166MHz MMX Pentium computer through a 28.8kbps modem connection!
Back in the days, you could see pictures loading line by line on your “browser” and Intel was promoting new processors on TV (see their 1997 funky advertisement below 😂)
Always has been, always will be…
Switching from being a teenager to a student, and then jumping into professional life, emails had always been part of the game. Over the years, I fine-tuned some technics to avoid being overwhelmed, even if you are receiving hundreds of them on a daily basis, as I do. Here are some guiding principles you might find…
I have just tested Lobe, Microsoft’s software aiming to make “Machine Learning” easy. How good is it?
Over the last months, Microsoft has issued many tools to simplify access to Machine Learning. To name a few:
- Azure Notebooks allows anyone to develop Jupyter notebooks in the cloud with a dedicated (but limited) computing power. This is an interesting alternative to what Kaggle or Google Colab (my favorite!) do offer.
- Azure Machine Learning offers auto-modeling features to help non-Data-Scientists explore their data with a guided platform and run their first Machine Learning models.
When I read a few days ago that Microsoft had released new software to “make Machine Learning easy”, I decided to give it a try and see what actually lies behind the announcement. …
Discover how the use of Scikit and Catboost models can help you deal with an unbalanced dataset and why SHAP is a great tool to explain A.I. predictions.
How two critical issues could be explored and solved at the same time?
- The first one is as old as employment is: people deciding to leave their employer for a better(?) job.
- The second has appeared over the last years: being able to explain - in an understandable way - how extremely complex A.I. models are making predictions.
To explore these two problems, we will need a couple of tools and data:
- As there is no real dataset publicly shared regarding employees’ resignation, we will use the one created by the IBM Data Science team to promote Watson. …
Discover why using “Functions” instead of “Linear Vectors” in Principal Components Analysis can help you better understand common trends and behaviors of time-series.
FPCA is traditionally implemented with R but the “FDASRSF” package from J. Derek Tucker will achieve similar (and even greater) results in Python.
If you have reached this page, you are probably familiar with PCA.
Principal Components Analysis is part of the Data Science exploration toolkit as it provides many benefits: reducing dimensions of a large dataset, preventing multi-collinearity, etc.
There are many articles out there that explain the benefits of PCA and, if needed, I suggest you to have a look at this one which summarizes my understanding of this methodology:
The intuition behind the “Functional” PCA
In a standard PCA process, we define Eigenvectors to convert the original dataset into a smaller one with fewer dimensions and for which most of the initial dataset variance is preserved (usually 90 or 95%). …
About
