Methods to compute distances

Posted on January 31, 2022

In computer science, it is often important to compute distances to assess similarity or dissimilarity between two points. Conceptually speaking, we could wonder how close are two cities. That is easy because we’re used to computing physical distances in km. However, how would you assess the similarity between a man in his 30s and a diagnostic of diabetes versus a woman in her 50s and a diagnostic of kidney disease? There are many different methods that I will list overRead More→

Worth mentioning IV: Web 3.0, a case for optimism, Bitcoin is not socialist, train models using genetic algorithms, and fully renewable energy may be an impossible dream

Posted on January 21, 2022

What Is Web 3.0 & Why It Matters: Web 3.0 is the current internet revolution. The main base is the decentralization of software and hardware accompanied by an increase in AI usage. The Case for Optimism: Kevin Kelly arguments in favour of generalized optimism. Aside from making you happier, the future requires some degree of ingenuity to embark ourselves in life-long projects. Why Bitcoin is not a socialist’s ally: Yanis Varoufakis argues that what makes bitcoin different from the fiatRead More→

How to make interpretable machine learning models

Posted on January 16, 2022

Generally, deep learning models are not interpretable. That poses a challenge especially when those models are intended for public use. Interpretable models can achieve the same performances providing better feedback to the developers and the users. Through an iterative process, the data fed into the model can be processed and tailored so that the number of data artifacts is reduced or even removed. With interpretable models, we also understand the way the prediction is made and remove the wrong interpretationsRead More→

New Year’s Resolutions 2022: Diverse and motivated

Posted on January 1, 2022

Because life is a never-ending learning process, every year I set myself new areas to improve. For many years I have been writing publicly my new year resolutions (2021,2020, 2019, etc) and my own assessments at the end of the year (2021,2020, 2019, etc). The process of writing them and after 365 days revisiting them is a practice that I find useful. This time I won’t split the objectives into different categories. Instead I’ll merge them together wihtout any lengthRead More→

2021 resolutions revised: PhD done, nothing else though

Posted on December 30, 2021

In Denmark, we started 2021 deep in lockdown. Things gradually improved early summer until we got an almost normal life back in December. This year was my last PhD year where I also submited my thesis (still need to defend). After reviewing the goals I can sum it up as lots of red. Personal Physical exercises: I kept doing regular workouts, specially runs but also bike rides and swims. This year I did my personal best times in semi-official races.Read More→

Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead (paper summary)

Posted on December 7, 2021

Deep learning models are usually regarded as black boxes. That is because they are not transparent about the way they reach the prediction. Humans cannot directly interpret the model with millions of parameters. Choosing ignorance can lead to unforeseen dangers. This is inherently a bad practice that should be minimized as much as possible. Current deep learning explainability tools aim to simplify the process to the outcome but does not really explain the “thinking process” that the model followed. ThisRead More→

Hidden technical debt in machine learning systems (paper summary)

Posted on November 30, 2021

Machine learning systems are wonderful. Many shapes and forms of machine learning algorithms are currently in use. Different models such as clustering like k-means, prediction methods like trees, or more advanced deep learning methods suffer from technical debt. In traditional software engineering, technical debt can be found in specific shapes. In addition to the “traditional” software engineering problems, machine learning systems also face new challenges. The following paragraphs present the different technical debt found in machine learning systems. 1. EncapsulationRead More→

Gain specific knowledge by making an observation

Posted on November 15, 2021

As defined by Naval Ravikant, specific Knowledge is the knowledge that you cannot be trained for it. It is located at the edge of knowledge and is very hard to figure out. To acquire specific knowledge requires a combination of existing tools but also creativity to combine them in innovative ways to create something new. To gain specific knowledge in your field you need to experiment. Everything starts with the question “what would happen if?“. 1. Make an observation Let’sRead More→

How to build assets?

Posted on November 2, 2021

Generally, when people talk about investing they think it’s about buying assets. However, that’s not the case. There are plenty of activities in which one can build assets with little to no money. If you create a blog, you’re creating an asset. If you learn, you’re improving your asset. When you go to work, you’re creating value. All of these activities have different pros and cons but one could say that all of them will bring you more money. SomeRead More→

Moving from power to talent

Posted on November 1, 2021

This is the most profound shift of all. Moving from the J.P. Morgan model of ambition to the Mark Zuckerberg model shifts the balance of power from capital to talent. Ambitious people have gone from writing cheques to writing code. Today the most ambitious individuals don’t own the means of production, if they can write code they are their own means of production (Marx would, perhaps, be surprised). This gives ambitious people unprecedented power. https://medium.com/entrepreneur-first/tech-entrepreneurship-and-the-disruption-of-ambition-4e6854121992