steppi

15 ms·

91.
To Explain or to Predict? (2010) [pdf](stat.berkeley.edu)
30 points by steppi 4y ago|6 comments
92.
by steppi 4y ago
There’s a nice 2010 article on this distinction by Galit Shmueli [0] called To explain or to predict? that explores this distinction in depth for anyone here interested in learning more. [0] https://www.stat.berkeley.edu/~
93.
by steppi 4y ago
Ah yes, got you. It is a false dichotomy because it neglects that there’s such a thing as Bayesian neural networks. Also, taking ensembles of ordinary neural networks with random initializations approximates Bayesian inference in a sense an
94.
by steppi 4y ago
The variational autoencoder is a Bayesian model. See [0] for instance. [0] https://jeffreyling.github.io/2018/01/09/vaes-are-bayesian.h...
95.
by steppi 4y ago
If one only had such an estimate for a single example, this would be true, but in aggregate over many predictions, the uncertainty bands can useful for decision making. This is an active area of research.
96.
by steppi 4y ago
This isn’t what your parent is saying. Many machine learning models are capable of producing calibrated probabilities. What Bayesian models give on top of this is that one doesn’t just predict a probability p, but a posterior distribution f
97.
by steppi 4y ago
Yes, there is a BayesianGaussianMixture class in sklearn.mixture [0]. There’s also sklearn.gaussian_process [1] which offers Gaussian process classification and regression, Bayesian learning algorithms which can be thought of as analogs of
98.
by steppi 4y ago
Not at the moment, but it’s a good suggestion. I should start putting something together like that. Feel free to email me though if you want to talk about books. You can find my email at the link in my profile.
99.
by steppi 4y ago
There’s a 1964 story, The Truth , by Stanislaw Lem on this matter which is worth a read [0] [0] https://thereader.mitpress.mit.edu/the-truth-by-stanislaw-le...
100.
by steppi 4y ago
I don’t disagree with your main point, but I’ve actually seen Einstein cited as someone who had a long scientific peak. He first transformed physics during the 1905 miracle year and then again in 1915 after completing the theory of general
101.
by steppi 4y ago
Got you. It seems like I'm the one misremembering. I thought they'd gotten more sophisticated than that by the end. I guess the story makes the ideas seem more powerful than they'd appear when laid out in a straight-forward w
102.
by steppi 4y ago
This problem actually comes up in The Goal . After the characters lift the bottlenecks in their factory, things go smoothly for a while but then they start to see starvation in other areas like you mentioned. The solution they come up with
103.
by steppi 4y ago
Very interesting. I wasn't aware of Pocklington. Some other early work in addition to the correspondence between Gödel and von Neumann mentioned in another comment: The French Mathematician Gabriel Lamé noted in 1845 that the Euclidean
104.
by steppi 4y ago
I think everything starting with the below paragraph of Smith’s article tries to give a decent answer to your question. As I said, I have surely provided more than enough introductory reading! Still, let’s ask: what has been published sin
105.
by steppi 4y ago
Working through all of Peter Smith’s suggested books from the original post should give one a solid understanding of much of what’s been going on in field up until fairly recently. Every recommended book in Smith’s list from the original po
106.
by steppi 4y ago
There’s an entire chapter devoted to Plato in Stewart Shapiro’s Thinking About Mathematics , the first book the author recommends. I think it’s pretty reasonable to recommend people start with an accessible contemporary survey rather than
107.
by steppi 4y ago
Early the following year, K. S. Srinivasan, a student at Madras Christian College who'd known Ramanujan back in Kumbakonam, dropped by to see him at Summer House. "Ramanujan," he said, "they call you a genius." Har
108.
by steppi 4y ago
For anyone who was subjected to the standard intro ordinary differential equations (ODEs) course and left unenlightened, I’d highly recommend Vladimir Arnold’s book on the subject [1]. It gives a lot of insight into the underlying geometry
109.
by steppi 4y ago
In a way, I think you and your parent are both right. The missing link is that the analogy is based on multivariate series. First order deals with each variable in isolation. X1, X2, X3, . . . Second order contains interaction terms like X1
110.
by steppi 4y ago
I thought it was uncontroversial that stochastic gradient descent can offer at least some protection against getting trapped in an undesirable local minima. Even in 2011 when I took my first machine learning course we were taught that while
111.
by steppi 4y ago
I'm very fond of the boost math documentation [0]. Also scikit-learn's documentation [1]. [0] https://www.boost.org/doc/libs/1_79_0/libs/math/doc/html/ind... [1] https:/&#x
112.
by steppi 4y ago
My understanding is that it's because extrapolation is difficult in general. Extrapolating the behavior of a periodic function like sine is one thing, but extrapolating the behavior of a 1d function that tends to infinity is another ch
113.
by steppi 4y ago
What a coincidence. Pyro has basically become my go to for Bayesian learning and I know those docs in and out. I’ve also skimmed that David Blei article before. Small world. I’m not very confident of success but it seems like an interesting
114.
by steppi 4y ago
It's a bit of a shame that neural networks are weak in this area because it would be incredible to have a good tool to approximate inverse functions in general. The fact that we almost never see neural networks being used as a tool fo
115.
by steppi 4y ago
Neural networks can approximate any function, but that doesn’t mean they do so efficiently. Depending on the function, they can require incredible amounts of neurons and training. At their worst, they devolve into a lookup table. It’s not
116.
by steppi 5y ago
It's easy to miss, but the second bullet point about "coding in a low-level language directly" links to another of the author's articles entitled Cython, Rust, and more: choosing a language for Python extensions [1] wh
117.
by steppi 5y ago
I published a paper in JOSS and it was by far the most pleasant and transparent journal submission and review process I’ve ever gone through. It felt like the future of academic publishing that I wish we could have distributed more widely t
118.
by steppi 5y ago
I have a bipolar diagnosis but with medication I've been symptom free for over 5 years now. For the past two years I've been taking 5mg Lithium Orotate nightly as a supplement to my main medication Lamotrigine [1]. I started takin
119.
by steppi 5y ago
William Daniel Hillis's dissertation about the Connection Machine is available online [1] and is a fascinating read. It generated some interesting past discussion here [2]. I started reading through the dissertation earlier this year w
120.
by steppi 5y ago
Right on. I'd actually edited out for brevity "perhaps a better analogy is that it's like saying Rubber Soul is the most famous Beatles album." and am now happy I left it to you to express better.

More ›