Title: Programming ML (page 200) - installing keras 2.3.1 doesn’t install TensorFlow, as the book suggests.
Thank you for the heads up, jert47! These libraries evolve at a steady pace. It will take a while to update the book, even the downloadable version–but I’ll update the setup instructions on progml.com soon.
Thank you again!
Hi, Paolo!
I solved this little problem by installing TF manually, version 2.7.0, which uninstalled Keras 2.3.1 and installed Keras 2.7.0.
As one might expect, it was just a minor inconvenience.
Otherwise, I must say that your book is amazing and your pedagogical talent is outstanding.
I’m a retired system programmer with a long history in telecom industry. So, all that ML hype did not even started during my active years. I made several attempts to conquer ML, but to no avail. You made it possible. Thank you.
Jerry T, Sweden
Carpe diem!!!
1 Like
Good to hear that the issue can be fixed easily, Jerry… and great to hear that you’re grokking machine learning! I have a similar experience: ML was quite far away from my daily experience for most of my career, and I’m not in my prime now… so it was a struggle for me to gather the resolve to study this stuff. Now I’m very glad that I put in the effort!
Hi again, Paolo!
Now I’m finished with “ML Programming”. And I’m still overwhelmed. Your book really explained to me what ML is (and what it’s not).
Your ability to formulate those advanced ideas in plain english is fantastic! And, I can only imagine the huge amount of work you
had to invest in the incredibly well-designed examples. This was the best book on any technical subject ever! And have red a few.
I am a little undecided where to go next on this amazing field. So, I can only wish you will write more on something I would consider
as exciting. ML on LEGO, perhaps? Now, with LEGO Robot Inventor and Raspberry Pi Build HAT …
Thank you and take care, for now.
Jerry Trymander, Sweden
Carpe diem!!!
1 Like
Jerry, thank you for the kind words! I read this first thing in the morning, and it really made my day. 
Leave a review on Amazon or Google, if you’re feeling like it–but no pressure!
Where next? Oh, there are so many routes that open up with this stuff. Right now I’m hooked on AI-generated art. Some creative minds are putting together public Colab notebooks (basically Jupyter Notebooks hosted on Google’s infrastructure) that tie together multiple open source models and algorithms to do some really impressive stuff. Try this one out for an example: Google Colab
But that’s just my thing, and you probably have your own ideas already. LEGO plus ML? That sounds amazing. (I’m a LEGO AFOL, as they call them these days 
.) By all means let me know where your explorations bring you!
This field moves at such a crazy pace. Some of the Colabs I’m playing with are based on ideas that were proposed in academic papers literally this year–and they have amazing concrete results. It’s all so exciting, honestly.
My question seems in line with this thread, so rather than starting a new one, I thought I’d ask it here:
@nusco
I am looking to work through this book as a part of a work-sponsored “professional development” and am trying to estimate the time I should set aside for this work.
I am relatively adept at Python, but have not done any ML programming to-date.
Based on that, do you have a rough recommendation for how much time I will need to work through all the exercises?
Heuristrocrat, I hesitate to give you an estimation, because the time needed for exercises is so subjective. I’m a very slow, methodical reader, so I know that a single chapter and the accompanying exercises might take as long as a couple of days to me–but I know many people who can read two or three of these chapters per day, including the exercises.
The code is not difficult for an experienced Python programmer. In fact, I aimed at making the code as simple as I could. The time-consuming part is understanding the concepts behind the code, plus changing the code and make experiments.
The book has 20 chapters over three parts, and the first two parts are probably the most conceptually dense–but the exercises in Part III are probably more time-consuming, because they involve programs that simply take longer to run. You can try going through the first chapter or two and extrapolate from there.
I understand, software estimation is hard in its own right. Your suggestion of trying out the first couple of chapters sounds like a sensible approach.
Thank you
1 Like