[HN Gopher] Pen and Paper Exercises in Machine Learning (2022)
       ___________________________________________________________________
        
       Pen and Paper Exercises in Machine Learning (2022)
        
       Author : ibobev
       Score  : 391 points
       Date   : 2025-03-21 20:07 UTC (1 days ago)
        
 (HTM) web link (arxiv.org)
 (TXT) w3m dump (arxiv.org)
        
       | antipaul wrote:
       | So who among current ML practitioners building "useful" ML could
       | solve some of these?
       | 
       | _Should they_ be able to?
        
         | jerrygenser wrote:
         | Nope, i don't think they should or need to be able to.
         | 
         | These exercises are useful for mathematical maturity which
         | results in intuition needed to develop novel algorithms or low
         | level optimizations.
         | 
         | Not needed to use existing train and deploy ML algorithms in
         | general.
        
         | danielmarkbruce wrote:
         | Depends on your definition of "ML practitioner", "building" and
         | "ML". Look at the section on optimization - some people have an
         | extremely good grasp of this and it helps them mentally iterate
         | through possible loss functions and possible ways to update
         | parameters and what can go wrong.
        
         | psyklic wrote:
         | Good news -- if you're not interested in extending state-of-
         | the-art and simply want to call APIs, you don't have to learn
         | ML deeply.
        
         | biotechbio wrote:
         | I am curious about the same thing. I worked as a ML engineer
         | for several years and have a couple of degrees in the field.
         | Skimming over the document, I recognized almost everything but
         | I would not be able to recall many of these topics if asked
         | without context, although at one time I might have been able
         | to.
         | 
         | What are others' general level of recall for this stuff? Am I a
         | charlatan who never was very good at math or is it just
         | expected that you will forget these things in time if you're
         | not using them regularly?
        
         | grandempire wrote:
         | Some people see studying as a chore and want to learn the
         | minimum to get the job done. Others find it insightful and fun
         | and enjoy doing problems and reading material.
         | 
         | Both approaches make contributions and can lead to success, but
         | in different ways.
        
       | kingkongjaffa wrote:
       | complete with solutions, beautiful, thank you for sharing!
       | 
       | I'd be interested in more of these pen and paper exercises, if
       | there is such a term, for other topics.
        
         | blackbear_ wrote:
         | Not sure which other topics you mean, but "1000 exercises in
         | probability" should keep you busy for a while (one can find the
         | PDF online). For other math oriented riddles, check out "The
         | colossal book of short puzzles and problems" and "The art and
         | craft of problem solving"
        
       | axpy906 wrote:
       | Love it.
        
       | simojo wrote:
       | Very neat! Reminds me of Tom Yeh's "AI By Hand" exercises [0].
       | 
       | [0] https://www.byhand.ai/
        
         | Sysreq2 wrote:
         | This is what I was expecting. Very much appreciated. OP's paper
         | is good - but I sort of feel like it's singing to the choir.
         | It's a great resource if you already know the material.
        
       | S4M wrote:
       | Looks neat! My only criticism would be that the solutions are
       | given right after the questions so I couldn't help to read the
       | answer of one question before thinking it through by myself.
        
       | plants wrote:
       | This is really neat! I work in machine learning but still feel
       | imposter syndrome with my foundations with math (specifically
       | linear algebra and matrix/tensor operations). Does anyone have
       | any more good resources for problem sets with an emphasis on deep
       | learning foundational skills? I find I learn best if I do a bit
       | of hands-on work every day (and if I can learn things from
       | multiple teachers' perspectives)
        
       | lucasoshiro wrote:
       | Seems to be cool, but, one of thing that most annoys me on
       | studying machine learning is that I may dive as deep as it is
       | possible in theory, but I can't see how it connects to the
       | practice, i. e. how it makes me choose the correct number of
       | neurons in a layer, how many layers, the activation functions, if
       | I should use a neural network or other techniques, and so on...
       | 
       | If someone have something explaining that I'll be grateful
        
         | joshdavham wrote:
         | I think you're just more interested in the practical side of ML
         | which is totally fine!
         | 
         | I'm a bit skeptical of how much math and theory the average MLE
         | actually needs. Obviously they do need some, but how much? I'm
         | not sure.
         | 
         | But on the other hand, the theoreticians often need much more
         | math. Something like the SVM could only have been invented by a
         | math genius like Vapnik.
        
           | nickpsecurity wrote:
           | Meat pro's told me linear algebra and some differential
           | calculus is the bare minimum. That's because some classes are
           | designed to build on only that. However, I think statistics
           | and probability would be helpful since they keep using
           | techniques from both. Also, you can solve many problems
           | without deep learning just using older, statistical methods.
        
           | lucasoshiro wrote:
           | > But on the other hand, the theoreticians often need much
           | more math
           | 
           | I see... I'm really not interested (at the moment, at least)
           | to be a pro, only to be able to train models for simple tasks
           | and understand the process
        
         | danielmarkbruce wrote:
         | Most things can't be learned via pure theory or pure practice.
         | Almost nothing related to work in the modern day can.
         | 
         | In ML not everything can be derived from theory. If it could,
         | we'd not have been so surprised by the performance of really
         | really large language models. At the same time, if you can't
         | reason about the math involved, you are going to have a
         | difficult time figuring why something isn't working or what
         | options you have - could be around architecture or loss
         | functions or choice of activation function or optimizer or
         | hyperparameters or training time/resources or a dozen other
         | things.
        
           | cubefox wrote:
           | > In ML not everything can be derived from theory.
           | 
           | And not every theory in ML has a lot of applications to
           | practice. For example, statistical learning theory has only
           | limited relevance in practice, and algorithmic learning
           | theory has basically none at all. There are a lot of
           | mathematical theories that are relatively old (often much
           | older than the deep learning boom and definitely older than
           | transformers) and that are more interesting from a conceptual
           | perspective rather than from the point of practical
           | applications.
        
         | hansvm wrote:
         | > all of the above
         | 
         | NFL says something about it being a wash for arbitrary data.
         | All results are going to be tuned to assumptions we have about
         | our data in particular (not too many discontinuities,
         | sufficiently well sampled, ...).
         | 
         | > neurons in a layer, how many layers, ...
         | 
         | Scaling laws are, currently, empirically derived. From those
         | you can pick your goals (e.g., at most $X and maximize
         | accuracy) and work backward to one or more optimal sets of
         | parameters. Except in very restricted domains or with other
         | strong assumptions I haven't seen anything giving you more than
         | that.
         | 
         | > activation functions
         | 
         | All of the above about how it can't matter for arbitrary data
         | and how parameters need to be empirically derived apply.
         | However: An important inductive bias a lot of practitioners use
         | is that every weight in the model should be roughly equally
         | important. There are other ways you choose activation
         | functions, especially in specialized domains, but when
         | designing a deep network one of the most important things you
         | can do is control the magnitude of information at each level of
         | backpropagation. If your activation function (and surrounding
         | infrastructure) approximately handles that problem then it's
         | probably good enough.
         | 
         | > neural network or other techniques
         | 
         | For almost every problem you're better off using something
         | other than a neural network (like catboost). I don't have any
         | good intuition for why that's the case. Test them both. That's
         | what the validation dataset is for.
         | 
         | > how it connects to the practice
         | 
         | For this article in particular, it doesn't connect to a ton of
         | what I personally do. I'm sure it resonates with someone. As
         | soon as pytorch or jax or whatever isn't good enough though and
         | you have to go implement stuff from scratch, you need a deep
         | dive in the theory you're implementing. To a lesser degree, if
         | you're interfacing with big frameworks nontrivially or working
         | around their limitations, you still need a deep understanding
         | of the things you're implementing.
         | 
         | Imagine, e.g., that you want all the modern ML tools in a world
         | where dynamic allocation, virtual functions, and all that
         | garbage aren't tractable. You can resoundedly beat every human
         | heuristic for phantom touchpad events in your mouse driver with
         | a tiny neural network, but you can't use pytorch to do it
         | without turning your laptop into a space heater.
         | 
         | Embedded devices aren't the only scenario where you might have
         | to venture off the beaten path. Much like the age-old argument
         | of importing a data structure vs writing your own, as soon as
         | you have requirements beyond what the library author provides
         | it's often worth it to do the whole thing on your own, and it
         | takes a firm theoretical foundation to do so swiftly and
         | correctly.
         | 
         | > how it connects to practice
         | 
         | That's a criticism I have of a lot of educational materials.
         | Connecting the dots is important in writing (competing with all
         | the advantages of brevity).
         | 
         | Pick on the Model-Based Learning section as an example. We're
         | asked, to start, to MLE a gaussian. (M)aximum (L)ikelihood
         | (E)stimation is an extremely important concept, and a lot of ML
         | practitioners throw it to the side.
         | 
         | Imagine, e.g., a 2-stage process where for each price bracket
         | you have a model reporting the likelihood of conversion and
         | then a second stage where you synthesize those predictions into
         | an optimal strategy. Common failure modes include (a)
         | mishandling variance, (b) assuming that MLE on each of the
         | models allows you to combine the mean/mode/... results into an
         | MLE composite action, (c) really an extension of [b], but if
         | you have the wrong loss function for your model(s) then they
         | aren't meaningfully combinable, ....
         | 
         | Something that should be obvious (predict conversion rates,
         | combine those rates to determine what you should do) has tons
         | of pitfalls if you don't holistically reason about the
         | composite process. That's perhaps a failure in the primitives
         | we use to construct those composite processes, but in today's
         | day and age it's still something you have to consider.
         | 
         | How does the book connect? I dunno. It looks more like a "kata"
         | (keep your fundamental skills sharp) than anything else. An
         | explicit connection to some real-world problem might make it
         | more tractable.
        
         | incognito124 wrote:
         | Maybe this will help:
         | 
         | https://github.com/google-research/tuning_playbook
        
           | danielscrubs wrote:
           | "Summary: When starting a new project, try to reuse a model
           | that already works."
           | 
           | "Summary: Start with the most popular optimizer for the type
           | of problem at hand."
           | 
           | This is like when game designers say: try a bunch of
           | different stuff!
           | 
           | Aka. we have no idea, but we have the manpower to throw
           | things at the wall for years. May the richest, not most
           | clever, company win!
        
           | lucasoshiro wrote:
           | Thanks!
        
         | moffkalast wrote:
         | > how it makes me choose the correct number of neurons in a
         | layer, how many layers, the activation function
         | 
         | Seeing massive ablation studies on each one of those in just
         | about every ML paper should be fairly indicative that nobody
         | knows shit about fuck when it comes to that. Just people trying
         | things out randomly and seeing what works, copying ideas from
         | each other resulting in some vague guidelines. It's the worst
         | field if you want things to be logical and explainable. It's
         | mostly labelling datasets, paying for compute and hoping for
         | the best.
        
           | bob1029 wrote:
           | > nobody knows shit about fuck when it comes to that
           | 
           | This is why I've abandoned neural networks as a computational
           | substrate for genetic programming experiments.
           | 
           | Tape-based UTMs may be extremely rigid in how they execute
           | instruction streams, but at least you can eventually
           | understand and describe everything that contributes to their
           | behavior.
           | 
           | Changing the fan-out from 12 to 15 in a NN is like ancient
           | voodoo ritual compared to realizing a program tape is
           | probably not long enough based upon rough entropy measures.
        
         | grandiego wrote:
         | Beginner here. A takeaway I got from the Andrew Ng's Coursera
         | course (specifically for neural networks) is that adding more
         | neurons and layers than the "minimum needed" is usually okay
         | (that is, no risk of overfitting when considering reasonable
         | regularization terms.) Sadly, there is no rule for that
         | minimum, so you must do trial and error; on the other side,
         | carelessly extending the network will be inefficient and
         | eventually slow. For the activation functions, the output
         | layer's is mostly determined by the problem being tackled, and
         | for the inner layers you usually start with ReLU and then try
         | some of the common variants using some heuristics (again
         | related to the current problem.) Of course you should consider
         | other successful models for similar problems as your starting
         | point.
        
         | godelski wrote:
         | Just curious, how "deep" have you gone into the theory? What
         | resources have you used? How strong is your math background?
         | 
         | Unfortunately a lot of the theory does require some heavy
         | mathematics, the type you won't see in a typical undergraduate
         | degree even for more math heavy subjects like physics. Topics
         | such as differential geometry, metric theory, set theory,
         | abstract algebra, and high dimensional statistics. But I do
         | promise that the theory helps and can build some very strong
         | intuition. It is also extremely important that you have a deep
         | understanding of what these mathematical operations are
         | _doing_. It does look like this exercise book is trying to
         | build that intuition, but I haven 't read it in depth. I can
         | say it is a good start, but only the very beginning of the
         | theory journey. There is a long road ahead beyond this.
         | > how it makes me choose the correct number of neurons in a
         | layer, how many layers,
         | 
         | Take a look at the Whitney embedding theorem. While this isn't
         | a precise answer, it'll help you gain some intuition about the
         | minimal number of parameters you need (and the VGG paper will
         | help you understand width vs depth). In a transformer, the MLP
         | layer post attention scales up 4x the dimensions before coming
         | down, which allows for untangling any knots in the data. While
         | 2x is the minimum, 4x creates a smoother landscape and so the
         | problem can be solved more easily. Some of this is discussed in
         | paper (Schaeffer, Miranda, and Koyejo) that counters the famous
         | Emergent Abilities paper by Wei et al. This _should_ be
         | discussed early on in ML courses when discussing problems like
         | XOR or the concentric circle. These problems are difficult
         | because in their natural dimension you cannot draw a hyperplane
         | discriminating them, but by increasing the dimensionality of
         | the problem you can. This fact is usually mentioned in intro ML
         | courses but I 'm not aware of one that contains more details
         | such as a discussion of the Whitney embedding theorem that
         | allow you to better generalize the concepts here.
         | > the activation functions
         | 
         | There's a very short video I like that visualizes Gelu[0], even
         | using the concentric circles! The channel has a lot of other
         | visualizations that will really benefit your intuition. You may
         | see where the differential geometry background can provide
         | benefits. Understanding how to manipulate manifolds is critical
         | to understanding what these networks are doing to the data.
         | Unfortunately these visualizations will not benefit you once
         | you scale beyond 3D as weird things happen in high dimensions,
         | even as low as 10[1]. A lot of visual intuition goes out the
         | window and this often leads people to either completely abandon
         | it or make erroneous assumptions (no, your friend cannot
         | visualize 4D objects[2,3] and that image you see of a tesseract
         | is quite misleading).
         | 
         | The activation functions provide non-linearity to the networks.
         | A key ingredient missing from the preceptron model. Remember
         | that with the universal approximation theorem you can
         | approximate any smooth, Lipschitz-continuious function, over a
         | closed boundary. You can, in simple cases, relate this to
         | Riemann Summation, but you are using smooth "bump functions"
         | instead of rectangles. I'm being fairly hand-wavy here on
         | purpose because this is not precise but there are relationships
         | to be found here. This is a HN comment, I have to overly
         | simplify. Also remember that a linear layer without an
         | activation can only perform Affine Transformations. That is,
         | after all, what a matrix multiplication is capable of (another
         | oversimplification).
         | 
         | The learning curve is quite steep and there's a big jump from
         | the common "it's just GMMs" or "it's just linear algebra" that
         | is commonly claimed[4]. There is a lot of depth here, and
         | unfortunately due to the hype there is a lot of stuff that says
         | "deep" or "advanced mathematics" but it is important to
         | remember that these terms are extremely relative. What is deep
         | to one person is shallow to another. But if it isn't going
         | beyond calculus, you are going to struggle, and I am extremely
         | empathetic to that. But again, I do promise that there is a lot
         | of insight to be gained by digging into the mathematics. There
         | is benefit to doing things the hard way. I won't try to
         | convince you that it is easy or that there isn't a lot of noise
         | surrounding the topic, because that'd be a lie. If it were
         | easy, ML systems wouldn't be "black boxes"![5]
         | 
         | I would also encourage you to learn some meta physics.
         | Something like Ian Hacking's representing and Intervening is a
         | good start. There are limitations to what can be understand
         | through experimentation alone, famously illustrated in Dyson's
         | recounting of then Fermi rejected his paper[6]. There is a
         | common misunderstanding of the saying "with 4 parameters I can
         | fit an elephant and with 5 I can make it wiggle its trunk." [6]
         | can help provide a better understanding to this, but we truly
         | do need to understand the limitation of empirical studies.
         | Science relies on the combination of empirical studies and
         | theory. They are no good without the other. This is because
         | science is about creating causal models, so one must be quite
         | careful and be extremely nuanced when doing any form of
         | evaluation. The subtle details can easily trick you.
         | 
         | [0] https://www.youtube.com/watch?v=uiB97cPEVxM
         | 
         | [1] https://www.penzba.co.uk/cgi-bin/PvsNP.py?SpikeySpheres
         | 
         | [2] https://www.youtube.com/shorts/_n7TMDnYdVY
         | 
         | [3] https://www.youtube.com/watch?v=FfiQBvcdFG0
         | 
         | [4] https://news.ycombinator.com/item?id=43418334
         | 
         | [5] I actually dislike this term. It is better to say that they
         | are opaque. A black box would imply that we have zero insights.
         | But in reality we can see everything going on inside, it is
         | just extremely difficult to interpret. We also do have _some_
         | understanding, so the interpretation isn 't impenetrable.
         | 
         | [6] https://www.youtube.com/watch?v=hV41QEKiMlM
        
           | Valk3_ wrote:
           | I wonder what kind of contributions can you make with a
           | strong math background versus someone with just undergrad
           | math background (engineer)? I know it's a vague question and
           | it's not so cut and dry, but I've lately been thinking about
           | theory vs practise, and feel a bit ambivalent towards theory
           | (even though I started with theory at first and loved it) and
           | also a bit lost, mostly due to the steep learning curve, i.e.
           | having to go beyond undergrad math (CS student with undergrad
           | math background). I guess it depends on what you want to do
           | in your career and what problems you are working on, but what
           | changed my view on theory was looking at other people with
           | little math background or with only undergrad math background
           | at most, that still were productive in creating useful
           | applications and or producing research papers in DL, which
           | showed to me that what is more important is having a strong
           | analytical mind, being a good engineer and being pragmatic.
           | With those qualities it feels like you can go top-down
           | approach when trying to fill in gaps in your knowledge, which
           | I guess is possible because DL is such an empirical field at
           | the moment.
           | 
           | So to me it feels like the "going beyond undergrad math"
           | formally is more if you want to be able to tackle the
           | theoretical problems of DL, in which case you need all the
           | help you can get from theory (perhaps not just math, but even
           | physics and other fields might help as well to view a problem
           | through more than one lens). IMO, it's like casting a wide
           | net, where the more you know the bigger the net is and hope
           | that something sticks. Going the math education route is a
           | safe way to expand this net.
        
             | la_fayette wrote:
             | I also wonder about that, e.g., considering the team behind
             | deepseek, was it more important for them to have great
             | engineering skills vs strong math backgrounds to achieve
             | this success?
        
           | lucasoshiro wrote:
           | Thanks for your time! Just added your commentary to my
           | favorites! :-)
        
         | porridgeraisin wrote:
         | [Edit] I seem to have turned this into somewhat of an
         | information dump...
         | 
         | Like other commenters said, you typically find those out by
         | just trying them out one by one and seeing what works. However,
         | you can prune the search space considerably given you know a
         | few things. These range from theory, to large experimental
         | results. For example, if google or someone widely deploys a
         | certain configuration, other people just use that. If large
         | experiments show that this and this setting for Adam works well
         | for NLP, other people just use that when working on NLP
         | problems. There was a large experiment done that showed that
         | the best activation functions were of the form alpha
         | _sigmoid(beta_ x). Sigmoid, tanh, Gelu, are all of this form.
         | Stuff like this is unfortunately the majority of the knowledge.
         | In fact, ReLU is being used without there even being a
         | universal approximation theorem[1] for networks using it! The
         | canonical one only works when sigmoids are used. No one cared,
         | because it worked in practice.
         | 
         | Typically, theoretical results are difficult to come by for
         | such a general model structure as neural networks. Think about
         | it, a theoretical result "for all neural networks" has very
         | little logical statements i.e constraints to work with, that
         | will then combine to produce other logical statements. So, you
         | would see theoretical results for a subset of architectures.
         | This is because the constraints that generate this subset give
         | us more to work with, and we can combine them in some way and
         | give a theorem or proof. Then, people find out empirically that
         | it works well for a more general network, too. An example of
         | this type of result is "dropout". The empirical motivation for
         | it was trying to train ensemble networks for cheap. In an
         | attempt to rest it on some theoretical grounding, it was shown
         | that for linear models it is equivalent to adding noise to the
         | input, which can be shown to be a good regularizer. But there
         | is no proof for more complex architectures. In practice, it
         | works anyway. But, you're not sure, so you include it in your
         | hyperparameter search.
         | 
         | There is some good theoretical grounding for many
         | regularization methods. My favorite is the proof that the very
         | straightforward L2 regularization on SGD, can be shown to
         | exactly limit the unimportant features, while not regularizing
         | much the important features. You can also search "stein's lemma
         | neural networks". I found [2], which is a talk on this topic,
         | and it is by Anima Anandkumar - always a good sign.
         | 
         | For activation functions, it is mostly that experimental result
         | that everyone relies on.
         | 
         | The universal approximation theorem [1] says that even a single
         | layer is enough to represent any function. However, there is a
         | practical difficulty in training these single-layer networks.
         | Deepening the network provides a lot of efficiency advantages.
         | Notably, for certain classes of functions, it provides an
         | exponential advantage (Eldan and Shamir 2016). There is a
         | wishy-washy(IMHO) theory called the Information Bottleneck
         | Theory, which tries to show that multiple layers stack on top
         | of each other, each uncovering one level of "heirarchy" in the
         | data distribution. This is seen in practice (see StyleNet) but
         | the theory is a little weak, again IMHO.
         | 
         | There is also a lot of tweaks done to the architecture in the
         | name of preventing the "Vanishing Gradients" problem - this is
         | a problem that arises because we use backpropagation to train
         | these networks. There is _some_ theory to help understand this,
         | that comes out of random matrix theory. But I don't know much
         | of it.
         | 
         | There is the old VC dimension theory of model complexity, but
         | that doesn't cleanly apply to neural networks as far as I have
         | seen.
         | 
         | [1] in case you are unaware, this is the theorem that makes
         | pursuing neural networks sound in the first place. It says that
         | you can always make a neural network that computes an arbitrary
         | function up to an arbitrary precision threshold.
         | 
         | [2] https://slideslive.com/38917864/role-of-steins-lemma-in-
         | guar...
        
         | informal007 wrote:
         | Some machine learning models are explainable, like tree model,
         | xgboost.
         | 
         | Neural network model are hard to explain, especially LLM.
        
         | yomritoyj wrote:
         | ML practice has for the moment far outstripped ML theory. But
         | even if ML theory catches up, the answers to your question will
         | get likely be still dependent on the nature of the process
         | generating the data and hence they would still have to be
         | answered empirically. I see the value of theory more in
         | providing a general conceptual framework. Just as the
         | asymptotic theory of algorithms today cannot tell you which
         | algorithm to use, but gives you some broad guidance.
        
           | xg15 wrote:
           | > _the answers to your question will get likely be still
           | dependent on the nature of the process generating the data
           | and hence they would still have to be answered empirically._
           | 
           | And I think that would be perfectly fine, or rather weird if
           | otherwise. Part(*) of the unpredictability of ML models stems
           | from the fact that the training data is unpredictable.
           | 
           | What is missing for me so far are more detailed explanations
           | _how_ the training data and task would influence specific
           | decisions in model architecture. So I wouldn 't expect a hard
           | answer in the sense of "always use this architecture or that
           | amount of neurons" but rather more insight what effects a
           | specific architecture would have on the model.
           | 
           | E.g. every ML 101 course teaches the difference between
           | single-layer and "multi"-layer (usually 2-layer) perceptrons:
           | Linear separability, XOR problem etc.
           | 
           | But I haven't seen a lot of resources about e.g. the
           | differences between 2- and 3-layer perceptrons, or 3- and
           | 32-layer, etc. Similarly, how are your model capabilities
           | influenced by the number of neurons _inside_ a layer, or for
           | convolutional layers, by parameters such as kernel
           | dimensions, stride dimensions, etc? Same for transformers:
           | What effects do embedding size, number of attention heads and
           | number of consecutive transformer layers have on the model 's
           | abilities? How do I determine good values?
           | 
           | I don't want absolute numbers here, but rather any kind of
           | understanding at all how to choose those numbers.
           | 
           | (There are some great answers in this thread already)
           | 
           | (* _part_ of it, not all. I 'm starting to get annoyed by the
           | "culture" of ML algorithm design that seems to love throwing
           | in additional sources of randomness and nondeterminism
           | whenever they don't have a good idea what to do otherwise:
           | Randomly shuffling/splitting the training data, random
           | initialization of weights, random neuron/layer dropouts,
           | random jumps during gradient descent, etc etc. All fine if
           | you only care about statistics and probability distributions,
           | but horrible if you want to debug a specific training setup
           | or understand why your model learned some specific behavior).
        
             | lucasoshiro wrote:
             | > every ML 101 course teaches the difference between
             | single-layer and "multi"-layer (usually 2-layer)
             | perceptrons: Linear separability, XOR problem etc.
             | 
             | Yeah, that's the point! ML related stuff seems to be
             | starting with simpler problems like linear separation and
             | XOR, then diving into some math, and soon it shows a
             | magical python code out of nowhere that solves a problem
             | (e.g. MNIST) and only that problem
        
         | sota_pop wrote:
         | The fact that it is still as much of an art as it is a science
         | means there is no "correct" values for these things. Only
         | heuristics and guess-and-checks.
        
         | SJC_Hacker wrote:
         | I guess the real question would be, given a budget of P
         | parameters, how many hidden layers in something like a multi-
         | layer perception is a good idea, and what are the size of those
         | hidden layers? As well as questions like is it every a good
         | idea to have a hidden layer that is larger than the previous
         | layer (including the input) ? Or are you just wasting compute /
         | parameter space?
         | 
         | I'm no expert, but a "rule of thumb" might be the more non-
         | linear the system is, the more hidden layers you would want.
         | 
         | Also let us consider the information in the input vector from
         | the perspective of compression.
         | 
         | How much you can compress without losing information depends on
         | the entropy of the system. Low entropy = high compression
         | ratio, while high entropy = low compression. High entropy is
         | essentially noise (total disorder), on the other hand very low
         | entropy just doesn't have much information (like a very long
         | string of 10101010 ...)
         | 
         | Most "interesting data" (video/audio/images) can be compressed
         | at ratios of about 50% before information loss kicks in. Note:
         | text can be compressed quite heavily, but that is partially
         | because the encoding is extremely inefficient - e.g. 8-bits per
         | char when really only ~5 are needed, and also of much lower
         | entropy (only ~30k words in the English language, for example)
         | 
         | On the other hand, information loss might not be such a bad
         | thing if the input data has extraneous information, which it
         | often does. This is why video, audio and image data can be
         | compressed at ratios 10x-20x before noticeable loss of quality.
         | 
         | So I think the answer would be, you don't want to decrease the
         | size of the previous layer, especially the input layer, by more
         | than about 10x-20x.
        
       | dang wrote:
       | Discussed at the time:
       | 
       |  _Pen and paper exercises in machine learning (2021)_ -
       | https://news.ycombinator.com/item?id=31913057 - June 2022 (55
       | comments)
        
       | imranq wrote:
       | If someone could turn these into an adaptive Khan Academy style
       | app, that would be incredible
        
         | mathandsurf wrote:
         | Just curious for you or anyone else, what would make such an
         | app compelling for you to use? And maybe not one that's just
         | aimed at learning the content of this document, but if you'd
         | like to think more broadly, an app aimed at helping you learn
         | and retain things that you're currently interested in,
         | studying, etc.
         | 
         | For these machine learning problems specifically, feel like
         | there are so many people that would greatly benefit from having
         | some form of spaced repetitive practice (as you mention like
         | the adaptive Khan Academy style app), or some other easy-to-use
         | format. I just wonder what other features people would want
         | that would make them want to use something like this over
         | learning with other resources (e.g., YouTube videos, reading
         | books, etc.)
        
           | imranq wrote:
           | There are already some resources like that like:
           | 
           | leetgpu.com
           | 
           | https://github.com/srush/GPU-Puzzles
           | 
           | For me its about a sense of progress, like in chess you can
           | have an ELO score. Or in Duolingo theres a roadmap. If there
           | were levels to this you could get more confident in your
           | abilities.
           | 
           | Right now the levels are basically bachelors, masters, and
           | PhD. Coarse and expensive
        
       | FilosofumRex wrote:
       | Funny how mathematicians always try to sneak their linear algebra
       | and matrix theory into ML. If you didn't know any better, you'd
       | think academicians had invented LLMs and are the experts to be
       | consulted with.
       | 
       | If anything academicians and theoreticians held ML back and
       | forced generations of grad students doing symbolic proofs, like
       | in this example, just because computational techniques were too
       | lowbrow for them.
        
         | nophunphil wrote:
         | Interesting. Can you share an example of this?
        
         | Abishek_Muthian wrote:
         | Interesting perspective, Would you have recommendations for
         | resources which prioritizes "computational techniques" over
         | "symbolic proofs"?
        
         | grandempire wrote:
         | If you want to contribute to ML and not just use existing
         | techniques, math skills are the most important limiter.
         | 
         | Who do you know making contributions who isn't fluent in linear
         | algebra?
         | 
         | Also why are you summarizing the entire field as "LLMs"?
        
           | Matthyze wrote:
           | Are math skills really? Most aspects of deep learning don't
           | require a deep understanding of mathematics to understand.
           | Backprop, convolution, attention, recurrent networks, skip
           | connections, GANs, RL, GNNs, etc. can all be stood with only
           | simple calculus and linear algebra.
           | 
           | I understand that the theoretical motivation for models is
           | often more math-heavy, but I'm skeptical that motivations
           | need always be mathematical in nature.
        
             | thecleaner wrote:
             | Every MLE who didnt study Math really likes to downplay its
             | importance. Yeah you dont need measure theoretic
             | probability, but you need a grasp of Lin Alg to structure
             | your computations better. Remember the normalization that
             | we do in attention ? That has a math justification. So I
             | guess yeah academics did have a role in building LLMs.
             | 
             | I mean computer scientists really do like to pretend like
             | they invented the whole field. Whereas in reality the
             | average OS, compilers, networks class has nothing to do
             | with core ML. But of course are also important and these
             | barbs dont get us anywhere.
        
               | Matthyze wrote:
               | I think you might've taken my point too strongly. Of
               | course math is very useful, and certain contributions are
               | purely mathematical. I just don't think it is as hard of
               | a requirement for innovation as was claimed.
        
             | anthk wrote:
             | Forget actual CS and proper engineering without discrete
             | math.
             | 
             | Also, without Shannon you wouldn't have neither Telecomms
             | nor Computer Science.
             | 
             | Heck, Lisp it's just a formalisation and implementation of
             | Lambda Calculus, which began as a paper... from a
             | Mathematician.
             | 
             | Also: https://hakmem.org
             | 
             | Forget any serious reading without Math skills.
        
             | grandempire wrote:
             | I'm not saying you can't use these existing techniques
             | without understand all the theory, but you're not going to
             | be able to find new techniques.
             | 
             | For example, how would you know optimizing a convolution
             | kernel is a good idea if you aren't familiar with linear
             | time invariant systems?
        
               | Matthyze wrote:
               | I think CNNs follow very naturally from the notion of
               | shift/spatial invariance of visual processing. That
               | doesn't require a mathematical understanding.
        
               | grandempire wrote:
               | Image processing and shift invariance come from DSP.
        
       | BeetleB wrote:
       | Isn't arxiv meant for research level papers? Surprised to see
       | this hosted there.
        
       ___________________________________________________________________
       (page generated 2025-03-22 23:02 UTC)