اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدA Farewell to Falsifiability
53
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
Some of the most obviously correct physical theories - namely string theory and the multiverse - make no testable predictions, leading many to question whether we should accept something as scientific even if it makes no testable predictions and hence is not refutable. However, some far-thinking physicists have proposed instead that we should give up on the notion of Falsifiability itself. We endorse this suggestion but think it does not go nearly far enough. We believe that we should also dispense with other outdated ideas, such as Fidelity, Frugality, Factuality and other F words. And we quote a lot of famous people to support this view.
قيم البحث
اقرأ أيضاً
This is an introduction of the volume of the journal Commentationes Mathematicae Universitatis Caroalinae dedicated to the memory of Vv{e}ra Trnkova}.
Cosmological models that invoke a multiverse - a collection of unobservable regions of space where conditions are very different from the region around us - are controversial, on the grounds that unobservable phenomena shouldnt play a crucial role in
legitimate scientific theories. I argue that the way we evaluate multiverse models is precisely the same as the way we evaluate any other models, on the basis of abduction, Bayesian inference, and empirical success. There is no scientifically respectable way to do cosmology without taking into account different possibilities for what the universe might be like outside our horizon. Multiverse theories are utterly conventionally scientific, even if evaluating them can be difficult in practice.
We propose a new way to implement an inflationary prior to a cosmological dataset that incorporates the inflationary observables at arbitrary order. This approach employs an exponential form for the Hubble parameter $H(phi)$ without taking the slow-r
oll approximation. At lowest non-trivial order, this $H(phi)$ has the unique property that it is the solution to the brachistochrone problem for inflation.
For seven decades, the widely held view has been that the formation, the migration and the decay of short-lived starspots explain the constantly changing light curves of chromospherically active stars. Our hypothesis is that these deceptive observed
light curves are interference of two real constant period light curves of long-lived starspots. The slow motion of these long-lived starspots with respect to each other causes the observed light curve changes. This hypothesis contradicts the current views of starspots. Therefore, we subject it to eight reproducible tests. Our new period finding method detects the two real light curves of FK Com. Our hypothesis is a total success: all real light curve parameters are directly connected to the long-lived starspots which are also seen in the Doppler images of FK Com.These parameters are spatially and temporally correlated just like in the Sun, including weak solar-like surface differential rotation. As for other chromospherically active stars, all eight reproducible tests also support our hypothesis. It explains many spurious phenomena: the rapid light curve changes, the short starspot life-times, the rapid rotation period changes, the active longitudes, the starspot migration, the period cycles, the amplitude cycles and the minimum epoch cycles. It also explains why the light curves and the Doppler images give contradicting surface differential rotation estimates even for the same individual star, as well as the abrupt 180 degrees shifts of activity (the flip-flop events) and the long-term mean light curves. We argue that the current views of starspots need to be revised.
The rapid recent progress in machine learning (ML) has raised a number of scientific questions that challenge the longstanding dogma of the field. One of the most important riddles is the good empirical generalization of overparameterized models. Ove
rparameterized models are excessively complex with respect to the size of the training dataset, which results in them perfectly fitting (i.e., interpolating) the training data, which is usually noisy. Such interpolation of noisy data is traditionally associated with detrimental overfitting, and yet a wide range of interpolating models -- from simple linear models to deep neural networks -- have recently been observed to generalize extremely well on fresh test data. Indeed, the recently discovered double descent phenomenon has revealed that highly overparameterized models often improve over the best underparameterized model in test performance. Understanding learning in this overparameterized regime requires new theory and foundational empirical studies, even for the simplest case of the linear model. The underpinnings of this understanding have been laid in very recent analyses of overparameterized linear regression and related statistical learning tasks, which resulted in precise analytic characterizations of double descent. This paper provides a succinct overview of this emerging theory of overparameterized ML (henceforth abbreviated as TOPML) that explains these recent findings through a statistical signal processing perspective. We emphasize the unique aspects that define the TOPML research area as a subfield of modern ML theory and outline interesting open questions that remain.
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
