Saar N: Fill & Download for Free

Because many of them were. Before Hitler, people were miserable. The most patriotic nation on the Continent had been humiliated in The Great War. It lost its overseas colonies. It had significant parcels of land (13%) forcibly removed (Alsace-Lorraine, N. Schleswig, parts of W. Prussia, Silesia as well as others). The Saar, technically part of Germany, was removed for international administration (rich in coal, to buckle any German war effort), and the Rhine was occupied and economically raided by outsiders. Hitler ended all of that. The NSDAP got the economy running again, gave people health care, vacations and filled their bellies. The Nazis rebuilt the proud German empire that had been --per Germans-- destroyed by foreign conspirators.Until the Autumn of 1944, with the fall of Aachen, the war was very distant for most. Even the quotidian bombing by the Allies did little to impact daily life for most Germans. No reliable opinion polls exist, and the heavily propagandized video footage can hardly be considered reliable. But despite that, the German factories were busy, food was on the tables, and people were (as best as can be ascertained) hopeful for their Führer to pull victory (if not a settlement on peace) out of his hat.

During my career, I always began to answer a call with “Vishwanath speaking” to head off the usual question Can I speak to … or Who is speaking please?Some others started with “Yes please?” or just “yes?”Among the Hindi speakers it was usually ‘haa~n jee’ or ‘boliye’ or ‘kahiye, kya baat hai?’ and I remember one who always started with ‘namaste jee’. The Telugu chaps said ‘cheppandi’ and the Kannada guys would say ‘hEli saar’.Even those who used Hello pronounced it in a variety of ways.Hallao, Halooo, HeLLLLo, haylo,Some used good morning, or good afternoon or good evening as applicable.A close relative, when answering his phone at home would always start with ‘Swami Sharanam’.During the early nineties, just before and after the Babri demolition, I had some BJP and RSS enthusiasts who started a trend and insisted on Jai Sri Ram! to begin their conversations. We likewise had some Ayyappa Bhakts who began with Swaamiye Sharanamayyappa!The boss was not amused and ticked them off and told them to take their Ram and Ayyappa Bhakti outside the office!—Edited to add:Shreyas Narasimhan rightly points out that ISKCON people use ‘Hare Krishna’

Note: This answer was given to a different question about why LASSO does feature selection with minor changes.The short answer to your question: We use a Laplacian prior because we believe most of our variables are not going to be useful..What do I mean by this? In a classical Bayesian regression (Ridge Regression) case, before we do anything (a priori) we assume that the conditional distribution of each [math] \beta[/math] given [math]\sigma^2 [/math]is normally distributed around zero. This distribution looks something like this:Where [math]Pr(\beta = 0) = 0 [/math]When we are doing a LASSO regression we are basically saying, look data, we like you, but we think you need to lose a couple … features. We like your tails [math]\beta[/math], but we want them a bit slimmer. Something kinda like this:Ideally, at zero this function is non-smooth, because we want [math]Pr(\beta = 0) \neq 0[/math] . Furthermore, we want the total region by zero to be larger so that the probability that [math]\beta[/math] is close to 0 is large.A good candidate for a distribution like this is the Laplace Distribution, also known as the Double Exponential Distribution.So now lets derive our model:Assume [math]Y|\beta,\sigma^2[/math] ~[math] N(\beta,\sigma^2)[/math].Further assume [math]\beta [/math]~[math] laplace(t,0) [/math]Then using bayes rule we know that and look at a single random variable [math]Y_i[/math] given observation[math] [/math] [math] [/math][math] x_i[/math]:[math]f(Y_i,\beta| \sigma^2, t,x_i) \propto f(Y_i | \beta, \sigma^2 ,t,x_i)p(\beta|t) [/math]Where [math]p[/math] is the prior density of the [math]\beta[/math] (the laplace) and [math]\propto[/math] means proportional to. Thus[math]f(Y_i,\beta| \sigma^2, t, x_i) \propto \frac{1}{\sqrt(2\pi \sigma^2)}e^{\frac{-(Y_i-x_i\beta)^2}{2\sigma^2}}\frac{1}{2t}e^{-\frac{|\beta|}{t}}[/math]Disregarding values that don’t interact with [math]\beta:[/math][math]-log(f(Y_i,\beta| \sigma^2, t)) = \frac{1}{2\sigma^2}(Y_i-x_i\beta)^2 + \frac{1}{t}|\beta| = (Y_i-x_i\beta)^2 + \lambda |\beta|[/math]Where [math]\lambda = 2\sigma^2/t[/math]We just derived the LASSO model!So the LASSO uses a Laplacian prior to encourage stricter feature selection and to express our prior belief that most of the variables are unimportant.