Hi there! Would appreciate recommendations for extra online resources (course material, videos, etc) that clearly explain limit theorems and convergence in probability and distribution!

In the CLT. If you repeatedly add individuals from the same distribution Y_t= X_1+X_2+...+X_t, then as time progresses their standardization will tend towards a standard Normal distribution. And in some sense the Normal distribution has maximum entropy.

In the H-Theorem. If you put some particles in a box, and a few assumptions. Then for any initial conditions, as time progresses their speed distribution will tend towards the Maxwell Speed Distribution, which also maximizes the entropy.

I'm not sure in what sense they have maximum entropy. But are these two theorems special cases of a larger theorem where we have convergence towards the distribution of maximum entropy?

Last Update: 2019/11/17

Foreword

A month ago I posted a model of how the Steins;Gate chronoverse works here. Since then, I have massively overhauled the model, including Steins;Gate 0 and more advanced mechanics. Enjoy.

 

Introduction

Ever since I first watched Steins;Gate, I knew I'd have to figure it out at some point. I've seen a lot of analyses about Steins;Gate. I mean, a lot of analyses. And they all converge on one point: a lot of questions about the plot are answered with "Just because" or "It's a plot hole". Ultimately, Time Travel in Steins;Gate is described as circular reasoning or nonsense.

So I wanted to see if it was possible to unify the rules for World Line Traversal into a cohesive, sensible, plot-hole-free framework. I know, I know, it's been 10 years since Steins;Gate came out, so there's not really much left to discuss about World Line traversal, right? I mean, no, it's not right. Come check it out.

 

Abstract (TL;DR)

This essay attempts to outline a model of Time Travel in Steins;Gate that accounts for and reconciles the majority of confusing and contradictory events. The driving idea is that Attractor Fields should not erupt spontaneously. Any event that plays causal significance to the Timeline must have actually happened at some point.

Events which lead directly to World Line shifts, World Line Shift Triggers, build up Causality. In general, the more World Line shifts that follow a specific Trigger Event, the greater the Causality buildup for that event. If enough World Line Shifts occur because of one specific set of such Trigger Events in particular, an Attractor Field is formed; Attractor Fields arise from a line of Causality. Rather than a bundle of strings, Attractor Fields can be better visualized as a chain, with each link being a subsequent World Line.

This Causality Buildup leads to 3 main phenomena collectively referred to as Convergence:

1. Trigger Events with high Causality in an Attractor Field become generally unpreventable from within the Attractor Field.

2. Unlike other high-Causality Trigger Events, the original Trigger event which led to the creation of the Attractor Field (the Attractor Field Trigger) is preventable. If it is prevented, it leads to Attractor Field Collapse, in which the World Line shifts back to the World Line from before t

Hi, everyone.
Probably very stupid question, I have been bumping into it regularly when reading articles dedicated to neural networks.

So is it just a fancy word for 'approaching minima'?

Thanks in advance.

I’ve been trying to beat the thing for hours, fought it about 4 times now. Nobody ever responds to my SOS either. I suck at wording stuff, but basically I need help. Any armor tips or charms or anything really. From what I’ve seen it’s called Xeno something? I’m on PS4 and I use the heavy Bowgun if that helps at all

Stuff I already know:

It’s looking like its head, tail, and front legs are it’s weak points. Basically shoot at whatever starts to glow

If it gets enraged I’m fucked

Abuse the ledges and crystals in the fighting area

Try to lure it into corners so you have room to dodge

Yea that’s about it. Anything would be appreciated, I can’t find other people shooting SOS flairs either

Sorry for the shitpost im just really desperate at this point lol

I'm simulating a steady state response of two restrictions in series. One of my goals is measuring the force on the face of a valve after the first restriction. What I'm seeing is that the goal never converges but oscillates within a reasonable range. The velocity plot makes it look like the flow after the first restriction is turbulent and dancing on the valve face.

1. Is there a way to view this animation of the evolution of a parameter (velocity in my case) through the couple hundred iterations I've ran? I'd guess no since I choose a steady state simulation, Solidworks probably isn't storing the state after each iteration.
2. What do you guys do if you can't get a good results in a full model due to turbulence or a lack of convergence? Is it valid to now model each restriction in isolation and combine their characteristic analytically in excel for example?

Did any other nation other than the earth kingdom had a sudden appearance of airbenders after HC? Imagine an airbender raised by firebenders or swamp benders, it would be a cool idea.

I get all the frustration and surprise, and I'm definitely not shouting "told you so", but the writing has been in Canonical's financial reports for the last few years, for everyone to see.

https://beta.companieshouse.gov.uk/company/06870835/filing-history

Mark Shuttleworth has always emphasized that these financial reports don't tell the whole story because you can't see what's going on inside its worldwide subsidiaries, but there are just a few important things you have to know, and those you could see:

• Canonical Group Ltd. was in increasing debt, 401 million dollars by March 2016.
• They made 103 million dollars in the fiscal year ending in March 2016 and still operated at a loss, while the debt increased substantially.
• They owe most of this money to Mark and a smaller amount to their own subsidiaries.
• Yes, people donate, but that's a joke compared to the amounts of money we're talking about.

If you account for the lost profits of the last 13 years plus the additional investments, Ubuntu Desktop and Convergence cost Mark Shuttleworth and the company about half a billion dollars. I don't fully follow the "Ubuntu Desktop paved the way for their success in the cloud and IoT". Their Cloud success is mostly bound to their work on OpenStack and containers (which 99% of the desktop users don't even know about), and Canonical giving away Ubuntu for free (if you spin up an Ubuntu instance on e.g. Microsoft Azure, it's free, but if you choose a Red Hat or SUSE instance you pay for a license). Their IoT success hasn't even materiallized yet and will be bound to Snappy, but the IoT crowd doesn't jump on Snappy because the Ubuntu Desktop is so cool, but because Snappy solves some actual problems other offerings don't.

I think the real issue here is that so many people, and it seems likely Mark Shuttleworth himself is one of them, chose to ignore this. People have been questioning Canonicals finances for years and got shot down all the time, often with obvious nonsense like "people donate". That's fine, it's Marks money, but it should have been clear to everybody that you can't just assume this will continue forever if you just don't care about the money. The desktop and mobile teams at Canonical employed more than a hundred people at times, that's about ten million dollars in staff costs alone. Per year. For multiple years. This investment either had to pay off on its own (e.g. through Desktop support contracts or mobile device sales), or by increasing re

https://preview.redd.it/n2xhdrhbl5d31.jpg?width=878&format=pjpg&auto=webp&s=4fdfb207393db5c43ce9210c3d24a2bd27e06bd4

Nearly all my models I've worked on (text generation or movie recommendations) in keras seem to reach their maximum accuracy, or their lowest loss within just a few epochs which sometimes takes a few minutes to an hour or so. The problem is that I've never had a model that generalized well to new data, or even its training data.

My question is, could models start to learn new patterns it failed to learn after stabilizing on a set loss or accuracy? Or once a model converges, do they generally lack the ability to learn anything from continued training? I'm curious about both the possibility of delayed learning, and whether it's a common occurrence. I've been too impatient to keep training for hours or even days, even though that's probably the only way successful models are trained.

Last Update: 2019/11/17

Foreword

A month ago I posted a model of how the Steins;Gate chronoverse works here. Since then, I have massively overhauled the model, including Steins;Gate 0 and more advanced mechanics. Enjoy.

 

Introduction

Ever since I first watched Steins;Gate, I knew I'd have to figure it out at some point. I've seen a lot of analyses about Steins;Gate. I mean, a lot of analyses. And they all converge on one point: a lot of questions about the plot are answered with "Just because" or "It's a plot hole". Ultimately, Time Travel in Steins;Gate is described as circular reasoning or nonsense.

So I wanted to see if it was possible to unify the rules for World Line Traversal into a cohesive, sensible, plot-hole-free framework. I know, I know, it's been 10 years since Steins;Gate came out, so there's not really much left to discuss about World Line traversal, right? I mean, no, it's not right. Come check it out.

 

Abstract (TL;DR)

This essay attempts to outline a model of Time Travel in Steins;Gate that accounts for and reconciles the majority of confusing and contradictory events. The driving idea is that Attractor Fields should not erupt spontaneously. Any event that plays causal significance to the Timeline must have actually happened at some point.

Events which lead directly to World Line shifts, World Line Shift Triggers, build up Causality. In general, the more World Line shifts that follow a specific Trigger Event, the greater the Causality buildup for that event. If enough World Line Shifts occur because of one specific set of such Trigger Events in particular, an Attractor Field is formed; Attractor Fields arise from a line of Causality. Rather than a bundle of strings, Attractor Fields can be better visualized as a chain, with each link being a subsequent World Line.

This Causality Buildup leads to 3 main phenomena collectively referred to as Convergence:

1. Trigger Events with high Causality in an Attractor Field become generally unpreventable from within the Attractor Field.

2. Unlike other high-Causality Trigger Events, the original Trigger event which led to the creation of the Attractor Field (the Attractor Field Trigger) is preventable. If it is prevented, it leads to Attractor Field Collapse, in which the World Line shifts back to the World Line from before t