Software

This page is part of the Hello Deep Learning series of blog posts. You are very welcome to improve this page via GitHub! In the previous chapter we taught a network of linear combinations and ‘rectified linear units’ to recognize handwritten digits reasonably successfully. But we already noted that the network would be sensitive to the exact location of pixels, and that it does not in any meaningful way “know” what a 7 looks like.

This page is part of the Hello Deep Learning series of blog posts. You are very welcome to improve this page via GitHub! In the previous chapter we described how automatic differentiation of the result of neural networks works. In the first and second chapters we designed and trained a one-layer neural network that could distinguish images of the digit 3 and the digit 7, and the network did so very well.

This page is part of the Hello Deep Learning series of blog posts. You are very welcome to improve this page via GitHub! In the previous chapter we configured a neural network and made it learn to distinguish between the digits 3 and 7. The learning turned out to consist of “twisting the knobs in the right direction”. Although simplistic, the results were pretty impressive. But, you might still be a bit underwhelmed - the network only distinguished between two digits.

This page is part of the Hello Deep Learning series of blog posts. You are very welcome to improve this page via GitHub! In this chapter we’re going to take the neural network we made earlier, but actually make it do some learning itself. And, oddly enough, this demonstration will again likely simultaneously make you wonder “is this all??” and also impress you by what even this trivial stuff can do.

This page is part of the Hello Deep Learning series of blog posts. You are very welcome to improve this page via GitHub! In this chapter we’re going to build our first neural network and take it for a spin. Weirdly, this demonstration will likely simultaneously make you wonder “is this all??” and also impress you by what even this trivial stuff can do. The first part of this chapter covers the theory, and shows no code.

This page is part of the Hello Deep Learning series of blog posts. Also, feel free to skip this intro and head straight for chapter 1 where the machine learning begins! Deep learning and ‘generative AI’ have now truly arrived. If this is a good thing very much remains to be seen. What is certain however is that these technologies will have a huge impact. Up to late 2022, I had unwisely derided the advances of deep learning as overhyped nonsense from people doing fake demos.

A from scratch GPU-free introduction to modern machine learning. Many tutorials exist already of course, but this one aims to really explain what is going on, from the ground up. Also, we’ll develop the demo until it is actually useful on real life data which you can supply yourself. Other documents start out from the (very impressive) PyTorch environment, or they attempt to math it up from first principles. Trying to understand deep learning via PyTorch is like trying to learn aerodynamics from flying an Airbus A380.

This is a living document - I’d normally spend a few days polishing everything, but since CRA talks are ongoing right now, there’s simply no time for that. Check back frequently for updates! Also please let me know urgently on bert@hubertnet.nl if you think I’m reading things incorrectly! As a follow-up to my earlier post on the EU Cyber Resilience Act, here I’d like to address some practicalities: how would it actually work.

First a round of thanks for the many people in industry and government who provided valuable links, background and insights! I could not have done this without your help! If you spot any mistakes, or have suggestions, please do contact me on bert@hubertnet.nl The EU’s new Cyber Resilience Act is admirable in its goal. And the EU is not alone in thinking something needs to be done about the dreadful state of security online – the Biden administration has just released its National Cybersecurity Strategy that has similar aims.

“The history of human thought would make it seem that there is difficulty in thinking of an idea even when all the facts are on the table” – Isaac Asimov In 2014, MIT’s Technology Review wrote a very interesting article about an attempt to have Isaac Asimov be part of a group of scientists attempting to think outside of the box. In this article they included a 1959 essay that Asimov wrote instead of continuing to taking part in this (classified) government work.

Deze reactie is als PDF ook te vinden op de site internetconsultatie.nl. Ook kan er tot en met 16 januari nog gereageerd worden. Aan de minister van Binnenlandse Zaken en Koninkrijksrelaties, H.G.J Bruins Slot en aan minister van Defensie, K.H. Ollongren Betreft: Tijdelijke wet onderzoeken AIVD en MIVD naar landen met een offensief cyberprogramma Geachte ministers, In aanvulling op de uitstekende reacties van mijn voormalige collega’s van de Toetsingscommissie Inzet Bevoegdheden, de stichting Bits of Freedom en het College voor de Rechten van de Mens wil ik graag het volgende kwijt.

Deze pagina is mede tot stand gekomen door kritisch commentaar & feedback van Marien Boonman en Hidde Brugmans & anonieme experts. Dank! En mochten er desondanks fouten zijn -> bert@hubertnet.nl We beginnen simpel, een woonhuis zonder zonnepanelen, verbruik minder dan 10.000kWh/jaar. Je elektriciteitsrekening bestaat (volgens de regels geldend op op 2 januari 2023) dan uit de volgende componenten: De kale kWh-prijs die je energieleverancier verzonnen heeft (waarover later meer, per kWh) Opslag Duurzame Energie- en Klimaattransitie (ODE) (3,05 cent per kWh in 2022, 0 cent in 2023) Energiebelasting (3,679 cent per kWh, 12,599 cent per kWh in 2023) BTW over al het bovenstaande (9% laatste helft 2022, per 1 januari 2023 weer 21%) Dit samen is wat je elektriciteitsbedrijf hun “tarief per kWh inclusief belastingen” moet noemen.

This is a refreshed & expanded copy of a very old page I hosted outside of this blog. I recently ran into “silent NaNs” again, and thought it might be a good idea to republish this advice here. A small post that documents something that almost no one appears to know. And if you do anything with floating point, you do need to know. Exceptions In C or C++, try this:

Dit artikel verscheen in iets andere vorm in de Volkskrant. Hier een versie met klikbare links ter onderbouwing: Recent presenteerde het kabinet een nieuwe wet die volgens de Autoriteit Persoonsgegevens inhoudt dat alle banktransacties van alle Nederlandse rekeninghouders gemonitord zullen worden in één gecentraliseerde database, met gebruik van algoritmes. De uitvoering zal in handen komen van een commerciële derde partij. Dit alles ter voorkoming van witwassen en terrorismefinanciering. Foto van Isaac Chou op Unsplash

This article is part of a series on (European) innovation and capabilities. Feedback is very welcome on bert@hubertnet.nl. I’d also like to thank the many proofreaders, but all mistakes remain mine! In the world of (high) technology, Europe is exceptionally weak at innovating. There are many ways to explain how this came to be, and a lot of the discussion focuses on unfair business conditions and regulation. In this post however I want to talk about important cultural and social reasons that are at least part of the reason, and present some possible solutions.

UPDATE: Nu ik niet meer werk voor de TIB kan deze uitleg van de nieuwe wet weer online. Via dit artikel zijn interviews en podcasts te vinden met meer context. UPDATE: Het rondetafelgesprek op 5 april 2023 in de Tweede Kamer is hier te bekijken en hier te lezen. En hier mijn spreektekst met voetnoten. Het kabinet heeft een wetsvoorstel gepubliceerd wat het toezicht en de bevoegdheden van de inlichtingen- en veiligheidsdiensten ingrijpend verandert.

A brief review of Stefan Zweig’s 1942 book “Die Welt von Gestern: Erinnerungen eines Europäers”, which I read in the earlier English translation by Benjamin W. Huebsch and Helmut Ripperger (public domain copy). There is also a more recent translation by Anthea Bell, The World of Yesterday: Memoirs of a European. It too appears to be excellent.   It is not often that one feels the urge to review an 80 year old work, but it left such a huge impression on me that I simply must tell you about what I’ve just read.

So in these wild times, the “ten year interest rate” is suddenly in the news again. But what does this mean? There is no committee that sets these rates, but it is also not something you can just measure. While you’ll find various other explanations of how to calculate the 10 year interest rate online, this may be the geekiest one, and definitely is the only one with a built-in calculator.

This is a page about some of the mechanics of ‘big data’, specifically how to store, transfer and process perhaps 100s of millions or billions of rows/events. Although some people think big data starts at 65k rows, at which point Excel used to lose it, I’m talking about gigabytes of data here. Infrastructure for this kind of thing abounds. Data can however come from lots of sources, and go to lots of different places.

I’ve seen some mediocre automated translations of my Dutch language resignation statement go round. To prevent any confusion, please find the story here in English: Until today I was one of the three members of the board that checks warrants for the Dutch intelligence and security services. This board is called “Toetsingscommissie Inzet Bevoegdheden” or TIB. If either of the civil or the military intelligence and security services of The Netherlands want to use their lawful intercept, SIGINT or hacking (& some other) legal powers, they have to first convince their own jurists, then their ministry and finally the TIB.