CBS Eye Recreated by Machine Learning

draft

Kids, dont try this at home.  These machine learning generative adversarial networks are for adults, you could hurt yourself. There are finally some decent results from a GAN on a pattern simple enough to work.  But this tells me almost nothing about what the technique can or cant do.  This test is very, very simple.

An example of the input images. 

And some of the generated images after training on about 125,000 different images.

GANs, Machine Learning & Learning on the Internet

draft

By using a simple test pattern and pushing through latest attempt at a GAN, I was able to confirm some suspicions I had about what was going on.  And through this I was able to use the right words in my Google search and come up with some posts and references to papers to help explain what was going on.

What I saw in my tests is that indeed my GAN was at various times doing a respectable job at recreating the image class.  And then it would seemingly lose its mind, or lose the thread, and go off into its own zone from which it would occassionally return.

The process has been productive and I have learned a lot, but the process is also suboptimal.  Its hard to tell from the Internet at first glance whether or not many of the posts that are thrown in your face have been written by people who actually understand what is going on.  My opinion at this point is that they do not.

In particular, what 98% of these posts about GANs did not tell me was that this very promising technique is not at all cut and dried.  That there are many questions and issues about pretty much every phase of the process, about what type of loss function, about what type of optimizer, about how to avoid "mode collapse" and many other critical issues.

So GANs are an exciting work in progress, a true research project, and unlike many other areas that someone from Computer Graphics might expect, nothing is cut and dried here.  It may work in one case but not another.

In my case, it truly makes progress towards the goal, but then it wanders off.

Here is a link to the best post I have found so far on this topic by Jonathan Hull.  He has written a dozen or so posts on GANs and so you might want to poke around and see more of what he has written.

https://medium.com/@jonathan_hui/gan-why-it-is-so-hard-to-train-generative-advisory-networks-819a86b3750b

Here is one of the input images as a reminder of what I am aiming for and then a few selected images of results that shows it is on the right track.  Remember, our GAN does not know a circle from a politician, so when you see something like a circle that is significant.

And here are some selected outputs from the GAN.

And then here are some images from when it has apparently lost its mind.

Test Patterns for Machine Learning

draft

I am slowly reducing my test patterns for machine learning to the simplest form that will tell me something.  

When this is a little further along, I may package them all up and put it on kaggle as a lesson for the others.

I should mention that these images have been through the Tensorflow/Keras data augmentation process, and are hence smashed in a variety of ways.  Here is an example of an original image as created via Nuke.  Still not perfect, but better.  For what I am doing, it really doesnt matter.

Shut the Government Down Now

draft

I am perfectly OK if they shut the government down until Trump/Pence is gone, the rapist is off the supreme court, and so forth. I will vote against any representative of mine who permits this travesty to go on even one more day. Its the moral equivalent of the hunger strike in the face of gross injustice.

Pence is Not an Existential Threat

draft

I have friends so terrified of a potential President Pence, that they are perfectly OK with Trump continuing in office and murdering women and children.  Thats how much they fear Pence.  I think they are out of their mind and need to get a grip.

Short of starting a nuclear war in order to encourage the end of the world and the return of Jesus the Christ child, there is nothing too bad that is likely to happen that could be any worse than Trump.  He cant get legislation passed.  He can put right wing nutty boys in the judicial system the way Trump does, but that is no worse.  Any executive order he signs that is egregious will be reviewed by non-Trump justices and can be overturned by the next administration.

And there is nearly zero chance that he can be elected president in 2020.  He has the charisma of an uninteresting rock.  He probably would not even be able to get the Republican nomination.

So yes, Pence for 1.5 years would be an embarrassment and a humiliation, but he does not represent an existential threat.

Running a More Controlled Test with Tensorflow/Keras

draft

After many interesting tests with Tensorflow/Keras, I am convinced I have no idea what is really going on.  So, I am doing what I should have done to begin with, which is to generate some test data and see how well the script i have written to implement a GAN works with known data.

I finally broke down and learned enough Nuke to be able to generate one of my standard test sequences.  512, 1K frames with two levels of noise and a 12 field chart.  

All the frames look like the one below, except of course that it animates, slowly.

Without Use, Knowledge Decays

draft

As part of the process of cleaning data for use in machine learning, I finally broke down and I am learning Nuke, that being the preferred compositing program after the predecessors (Shake & Chalice) were destroyed by Apple out of greed. 

But the point is that with the passage of a few years, all kinds of skills and knowledge go away and I am stuck with the frustrating process of trying to remember what used to be second nature.

Let this be a warning to youth out there.  This can and will happen to you.

Why Impeachment is Not About Removing Trump (at least not exactly)

draft

This one of those essays that require at least a day to write, and probably several days to write, assuming one wants to do it well.  But I dont have that time so I will just have to sketch out the ideas here and hope you will be indulgent.

The question is as follows.  Is it necessary or desirable to impeach Donald Trump even if it is impossible to convict him because Republicans control the Senate and Republicans are traitors?

Some would say it is not desirable because it would piss off many good Americans who will think that the Democrats are being vindicative.  Others would say it is not desirable because it probably wont remove Trump from office, but it *might* and then we have Pence to deal with.  Still others would say it is not desirable because as far as they can tell Trump has not been "proven" to have committed a crime.

I think it is desirable, even required, and for the following reasons.

First, there are no good Americans who would be pissed off.  The only Americans who support Trump at this point (maybe not initially) are traitors who want to destroy this country, put rapists on the supreme court, etc.

Second, the only mechanisms we have to remove a madman like Trump is impeachment (or article 4 of the 25th Amendment).  If that means Pence, that means Pence.  Pence is a religious nut will try to put gay people in extermination camps, but he will fail.

Third, it is not necessary to prove Trump committed a crime to impeach him.

But the most important reason is this.

The Republicans have tried to destroy this country and we have to fight back.  Now, and forever, now or not at all.

Machine Learning and Nebulae

draft
Damn this machine learning can be fun! The images are definitely getting better and I think it would be petty to complain that the "generated" images dont look much like the input "real/training" images. That would be the concern of small minded people and everyone here I am sure operates on a higher level. Oh yes, also, the "metric" of good vs bad (error, incorrectness, what have you) also seems to have no relationship whatsoever to the images being "judged". 

Again, we must rise above such nonsense!

Nebulae, who needs nebulae?

Noteworthy Brief from the FBI about Cohen

draft

I have not read the brief that Mueller has filed but I have read about it and seen a few excerpts that seemed so noteworthy as to bring it to the attention of my friends.  In this paragraph, the Special Council's office makes the following point about Michael Cohen, esq.

While many Americans who desired a particular outcome to the election knocked on doors, toiled at phone banks, or found any number of other legal ways to make their voices heard, Cohen sought to influence the election from the shadows. He did so by orchestrating secret and illegal payments to silence two women who otherwise would have made public their alleged extramarital affairs with Individual-1. In the process, Cohen deceived the voting public by hiding alleged facts that he believed would have had a substantial effect on the election.