462 karma · joined June 13, 2020
Also, debugging colors on a 3D surface is not an easy task (debugging in general in 3D is not easy). So if the rendering is nice and seems correct, you tend to think it is.
And if it was not, and you didn't encounter something that bothers you, it doesn't matter that much, after all, what is important is the final rendering style, not that everything is perfectly accurate.
Honestly, the language became kinda harsh for newcomers, what we see as developpers is 'it's like pseudocode that runs'.
But a beginner is often left behind the billions of methods in each class. He is not used to documentation, and spend quite a huge amount of time learning by heart stupid things like 'len()' in this case it's '.len()' here it's '.length',etc.. For many meany methods that all have their idiosyncracies.
At least in c/(easy)c++, you need to build yourself most of it, helping the understanding.
I'm not completely against python as a first language, but it need to be teached well, and that could include working with a very minimal set of functions on every objects. Then you can expand and incorporate more and more methods that make life easier.
It's nice to know I'm not the only one thinking about that.
The trick for me is that it's a path in a graph for each student, so even if some component is not as strong for one student, he can fill the gap by taking another route. A good framework would be resilient if it finds many possible paths to reach the same result, and not forcing one path. But then, teaching in this way is more difficult.
A post-Processor transform the proof into lean (with any official lib loaded). It is automatically verified. If something is missing, the post processor ask to write in English the missing parts. Iterate like this.
The lean proof is hidden in the final document, and can be displayed if needed. Or even, we get an English version that can be easily retransformed into lean at will.
Bonus point: we can query the document to give more details on part of the proof and it outputs (expanded) lean formatted as nice English.
Note: there is no need to have all the math self contained in the document, he can say to assume some theorem true to do his proof. And this would be reflected in English.
I work at the university and we build acquisition systems with exotic cameras and screens, do you think we could meet one time to discuss possible (commercial and research) projects ?
Is it effective to teach? Or just easier to teach?
In my education we copied the black board by hand, and that allowed us to learn how to summarize information. Writing by hand also improved memory. And seeing the teacher thinking and choosing how to present the information in real time had a lot of value.
Further I find it way more difficult to make a slide presentation that effectively teach something than only having a blackboard that forces us to really explain the steps.
At uni I teach with slides, but they are done in a way that mimics blackboard writing. It takes an enormous amount of time to make them, but the students are very happy.
What are your opinions?
Do you happen to still have the homework questions/solutions somewhere ?
At the moment the new ones are locked behind CMU credentials.
In the article they say that parabolic math was not available in 1999 computer games, but the bio rifle seems to do it just fine, even at the time ;)
For the ones that want to have fun with the maths, this demo is awesome https://youtube.com/shorts/akEGxm5rcC4
But this video also means, that you still aim at the player head when he is falling straight, you will just hit him a little lower than expected ;)
After that, I can't recommend enough Bishop, machine learning, and the Bishop on deep learning (books)
Probability https://youtube.com/playlist?list=PLoROMvodv4rOpr_A7B9SriE_i...
Basic algebra and calculus https://tutorial.math.lamar.edu/
Real analysis https://youtube.com/playlist?list=PL0E754696F72137EC
All those are really basics and start from (almost) nothing
>> for the leisurely researcher, self-study must include the discipline’s foundational texts
It completely depends on the depth at which you are researching for leisure. If you like to do it to learn some stuff, just using LLMS, wiki and books is perfectly fine!
If you're trying to make a breakthrough in human understanding, it is useful to understand very very well the foundations. It is not required, sometime, another background is enough to see clearly some research issue.
But as you're doing research for fun, if you want to push the boundaries, having very strong foundational knowledge is awesome and very empowering.
Take for example the LLMs, you have many levels of comprehension that you may aim for (just an example):
- simply using occasionally (existance awareness)
- experimenting with prompts engineering (exploratory research)
- studying the neural architectures (understanding better what's going on)
- understanding how the architectures are validated (if they are...)
- finding limiting cases (deeper exploratory research)
- thinking about in under the perspective of bayesian/classical statistics (research at the normal mathematical/engineering level)
- going deep into characterizing the architectures in terms of manifolds, or any other suitable structure that you believe is true (deep fundamental research)
- opening up new questions about what can be learned and why (leading to another research field)
The first 3 are available inside a LLM, on Wikipedia and in books. The rest you need to fetch some papers. The more you go down the path, the more obscure the papers gets.
Up to the end, the last point, where you land in survey papers for the other field, and then you can start again with new questions. The good thing is that most of the fundamental research can be readily applied to understand the new field! It's like a jump start.
But at the end of the day, depending on which depth you want to do your leisure research, the deeper you go, the more foundation papers become important, not just from their content, but also the powerful ideas that can be transferred in other research projects.
Maybe.. One thing that is missing from what I say is that it's not about the quantity, or even the quality of the material you read. But it depends on engaging in deep thinking about it! Someone can spend his life reading and not become smarter, maybe only more knowledgeable.
I my opinion, deep thinking is maybe the most fun and useful skill in leisure research, as in paid research.
Peter Van Roy, Concepts, Techniques, and Models of Computer Programming, 2004
It uses the language developed in this book (Oz).
The book constructs many of the current programming paradigms, starting from almost nothing, with an emphasis on software design.
See the paradigms poster built http://www.info.ucl.ac.be/people/PVR/paradigmsDIAGRAMeng201....
I think finding someone that guide you through the book allows to go way faster than alone.
Now, while the concepts were extremely illuminating, I think following the book but doing the exercises in haskell is also a nice path. Haskell was another lecture following the one on SICP by the same teacher, and it uses most of the same ideas, sometimes formulated a little differently, but on a more modern language.
I still remember my exam where we received ~6 pages double column of printed lisp, no comments. The code was a compiler, and we needed to change it so that it can jump back in time using call with cc. Difficult to do in 3h and on paper but very fun.
I don't remember exercises, but I expect they are as long as SICP, while maybe there is more emphasis on learning theory more than practicing coding
Nice idea, but I'm also planning a book on advanced cuda programming, and I will cover convolutions there, I'll think if it is interesting to have some kind of cross chapter that is valid for both books.
> Noises I cover those in normal mapping
> anisotropy I have a full chapter on brdf that includes anysotropy for raster and raytracing, I use it also to derive anisotropic filtering
> Color spaces I have to make choices, and I don't think I'll cover in depth color spaces, it's a topic I know, but I find it very annoying to write about, in particular, making it fun to read is difficult
> acceleration structures I will have a short section on them as I use them for raytracing, at the moment I use only a basic octree, but I could expand to include other partitioning structures if that does not make the book too heavy to read
> diffusion processes and 2d/3d solvers Will be better in the cuda book
> complex numbers I derive complex, quaternions, and geometric algebra for rotations
> orbit based fractals I'm not familiar with this topic, apart from basic fractal rendering, I'll look into it You may find this paper fun to read https://arxiv.org/abs/2402.06184 even if it's not exactly about their rendering
> metropolis sampling I do not expect to write about mcmc and metropolis in this book, but I keep the idea
> Demoscene I need to ask a friend that worked in the Demoscene if he would be interested in write something, but I'm not the right person to write about it as I only have a summary knowledge of it, I'm afraid that will be out of scope
Thanks very much for all the ideas!
I already have a full chapter on aliasing.
The exercices use renderdoc and nvidia tools ;)
There is an even greater beauty than complexity when very complex subjects can be reduced to simpler ones without losing their power, it requires a greater understanding of complexity than what is usually seen.
As an illustrative example : QED: The Strange Theory of Light and Matter, Feynman
I do this at least from 10 years without bad effects, the positive effects are that I'm more awake and feel fresh longer, in control of my mental power.
It is quite easy to do it while having normal life obligations, I can wake up sooner if something needs to be done, and then, after the task, I can continue to do my day until sleepy, at which point I can sleep again. So, if I have many meeting on a day, I can stay up all day without issues, maybe I'll sleep more in the evening or following day.
I think the quantity of sleep I require depends on the tasks I'm doing. If some task is particularly mentally exhausting I'll automatically do more naps during the day, this in turn makes me more efficient at solving difficult problems.
It is extremely comfortable to solve hard problems in this way, when you work too long on it, you get sleepy, and as you wake up, you are again fully focused and with fresh perspectives.
It's all that you learn doing it, it is not just knowledge, it's the scientific method, social skills, learning to communicate science.
A PhD is a personal path, an experience in life that changes it profoundly, I hope that those students will be able to handle any kind of future issues, not just a very niche applicative aspect of deep learning.