Learning Deep Learning with Python

In brief moments of boredom it is easy to get overambitious and overcommit to some exciting project. On that topic, I bought the book Deep Learning with Python written by François Chollet to have something to enjoy myself with during this holiday season. With the public holidays now coming to an end, it seems that my free time flew by in a flash—book in hand or not. While my plan to learn a bit about deep learning only resulted in reading the book’s first chapter it was a great start and I have the rest of the year to finish what I started. With that, let me give you some inspiration as to why deep learning is the thing to learn for 2022.

Why learn about Deep Learning?

When my thirsts for knowledge gets triggered I go on the hunt for knowledge, and I don’t want to go for any kind of knowledge—I want the knowledge to be meaningful.

My search for meaningful knowledge began with me going back to read my old post about Øredev from 2019 to see what I found interesting back then but never really picked up. Keras caught my eye, and so it was decided that I should learn about Deep Learning.

There is also another good reason for learning deep learning: At my job at Sectra we are regularly treated with university students that write their master’s thesis in collaboration with us. It’s great fun for everyone involved, at least up until the day of thesis presentations. On that day, the bright students talk about what they’ve been doing for the last 5 months and as soon they start speaking their machine learning lingo (they almost always do) I’m disconnected from the information because I simply don’t understand the language that has become my generation’s knoparmoj. It annoys me, and the only remedy is to learn the terminology. So thank you, thesis students, for pushing me into learning an exciting and relevant topic.

Key takeaways from Deep Learning with Python (2nd ed)

Following are some short excerpts from the book’s first chapter that I found intriguing.

[…] machine learning, and especially deep learning, exhibits comparatively little mathematical theory—maybe too little—and is fundamentally an engineering discipline

As much as I enjoy mathematics, hearing that deep learning as a practice doesn’t require extensive mathematical knowledge is super welcome. It’s kind of like cryptography, just because I enjoy putting some AES encryption in place doesn’t mean I feel the need to know exactly how the bits are XORed into secrecy. The same things can probably be said about those interested in cryptocurrency. Crypto junkies and Keras connoisseurs alike don’t want or need a perfect understanding of the mathematical theory, they just need to understand the tools well enough to use them and get rich!

The term “neural network” refers to neurobiology, but although some of the central concepts in deep learning were developed in part by drawing inspiration from our understanding of the brain (in particular, the visual cortex), deep learning models are not models of the brain. There’s no evidence that the brain implements anything like the learning mechanisms used in modern deep learning models.

Thanks to my fascination with anatomy and physiology I always thought that neural networks are these super cool things that require a PhD in computer science as well as an M.D. to fully understand. Turns out that neural networks have little in common with biology. Bit of a false advertising perhaps, but naming things are hard. Should I have named them, I would have went for tranche networks to draw some inspiration from the financial world. A missed opportunity in my opinion, but perhaps the machine learning people weren’t quite so Francophile. My conclusion from this is that economists name things to make them inconceivable to normal people while computer scientists name things in order to make them sound cooler than they are.

We’re still exploring the full extent of what deep learning can do. We’ve started applying it with great success to a wide variety of problems that were thought to be impossible to solve just a few years ago—[…], assisting oncologists or radiologists with interpreting medical imaging data, […].

The author does a good job in bringing some sensibility onto the AI hype train and tone the hype down a notch, while also presenting the actual achievements. If you want to not only join the hype train but also make it go faster then Sectra might have a chair in the driver’s seat just for you!

In a not-so-distant future, AI will be your assistant, even your friend; it will answer your questions, help educate your kids, and watch over your health. It will deliver your groceries to your door and drive you from point A to point B. It will be your interface to an increasingly complex and information-intensive world. And, even more important, AI will help humanity as a whole move forward, by assisting human scientists in new breakthrough discoveries across all scientific fields, from genomics to mathematics.

Well, if this doesn’t hype you up about AI then I’m out of options, but if you do want to learn more about machine learning and deep learning and even read a book about it then this might be your lucky day!

Book Giveaway

When I bought the book I received a free copy for a friend or two. So if you’re interested in the book and want your own copy, it would make my day if you used one the following links:

• Best gift: Deep Learning with Python, 2nd ed
  
• Second best gift: Deep Learning with Python, 1st ed

There’s no affiliation or such, I just hope for these books to find a couple of new owners because they’re good books! The offer ends at January 15, 2022 but hopefully this will be sorted by then. If you claim the book, feel free to send me a message to let me know. And if the links don’t work, let me know as well and I’ll deal with it appropriately.

Happy new year and happy learning!