Category Archives: Machine Learning

TMDNE

TMDNE: This Mathematician Does Not Exist

Image Generation, Machine Learning, , , , , , , , , ,

It’s like the math is the seeds or DNA and the code is the fertile ground and the sun shines, or in our case electricity flows to CPUs and GPUs.

Ever pondered the possibilities when algorithms dabble in the divine art of mathematics? Prepare yourselves for TMDNE – where Stable Diffusion meets the blackboard of the mind! 🧮🤯

Inspired by the brilliance of projects like “This Waifu Does Not Exist” (TWDNE) by Gwern Branwen, TMDNE takes a quantum leap into the realm of imaginary mathematicians. These aren’t your conventional number crunchers; they’re the brainchildren of algorithms, immersed in the ethereal dance of imaginary theorems and proofs. 🎨➗

Much like TWDNE introduced us to AI-generated characters that captured our otaku hearts, TMDNE invites you to witness the magic of machine learning as it conjures up mathematical maestros who never attended a single conference. 🤓📚

Get ready for a symphony of computational creativity, where algorithms showcase their virtuosity in crafting mathematicians who not only ponder the mysteries of prime numbers but might just solve the Riemann Hypothesis over a cup of virtual coffee. ☕📏

Embark on this intellectual odyssey where Stable Diffusion meets the abstract canvas of mathematical imagination, proving that in the world of artificial intelligence, even mathematicians can be a product of algorithmic brilliance! 🔢🤖

 

Grid Search in Stable Diffusion

The images in the following video were created by doing a grid search with Stable Diffusion, with slight variations of words in the prompt and scanning the guidance scale. The seed value remained constant. The images were also audited manually for obvious artifacts and flaws.

Finally a slideshow was created by using the following ffmpeg command:

ffmpeg -framerate 1/3 -pattern_type glob -i "image\([0-9]*\).png" -c:v libx264 -r 30 -pix_fmt yuv420p slideshow.mp4

This results in a slide show with a change every three seconds on the images.

The Idea

Math -> AI -> Imaginary Mathematicians

OK, AI is math really, a bunch of chain rule linkage and matrix multiplies for the most part and to make it do something useful the math gets worked into code as the substrate. It’s like the math is the seeds or DNA and the code is the fertile ground and the sun shines, or in our case electricity flows to CPUs and GPUs. In 2024, it’s gotten really good. imagine generation that is. I did some AI image modification years ago but, Stable Diffusion has gone way far, fast. So wouldn’t it be interesting and humorous to generate images of math being worked on, by a mathematician generated by AI. As can be seen from the images, this points out a weakness too. People, it kind of nails it, backgrounds too. Text and numbers, not so much. I’ll be interested in seeing it nail this too, only a matter of time.

Grid Search & Ground Rules

As an AI researcher, I am always interested in pushing the boundaries and grid search is something that is a normal event for me. I’ve played with Stable Diffusion for quite a while and I was interested in seeing what a semi-serious grid search would reveal now that it is seriously capable of generating graphics. I decided to stick with images that are not ‘fantasy’ or too enhanced to keep them real looking so that they are easier to judge.Keeping the ‘look’ in the same zone also helped keep some degree of uniformity but, Stable Diffusion varies the image within a range that seems to work well.

Yet another sidequest

This started as a bit of a sidequest as I am working on a project called TinyMath that is all about identifying how small a GPT model can get and get a passing grade for math. This way it is possible to score objectively and see how model size and loss stack up against the ground truth of taking an ‘exam’. I just happened to create an image using Stable Diffusion of Misato Katsuragi from Neon Genesis Evangelion at a blackboard working out some math when I was pondering this, kind of a nod to TWDNE on a document I wrote on makemore. That combined with an image that I saw in 2020 when my partner Renee was searching for a new hair style, was another inspiration. The image below only shows a partial view of the woman and I often wondered if the other views were out there somewhere, or could AI come close to generating the other views.

The Two Images that got TMDNE going

Misato Katsuragi as a Math Teacher
Misato Katsuragi as a Math Teacher
Layered Stacked Bob for thick hair. Great style and color.
Layered Stacked Bob for thick hair. Great style and color.

 

Notes & References

TWDNE

“This Waifu Does Not Exist” (TWDNE) is a project created by Gwern Branwen. It involves the generation of images of fictional characters using Generative Adversarial Networks (GANs). GANs are a type of artificial intelligence algorithm that can generate realistic-looking images based on patterns learned from a dataset.

In the case of TWDNE, the GANs are trained on a dataset of anime-style characters, and the output is entirely computer-generated images of characters that do not exist in reality. The term “waifu” is a romanization of the Japanese word for “wife” and is often used in fandom to refer to fictional characters, especially those from anime and manga, for whom someone has a strong emotional attachment.

It’s worth noting that the images generated by TWDNE are not real individuals, but rather creations of the algorithm based on the patterns it learned during training. The project is an interesting example of the capabilities of AI in generating realistic-looking content.

Prompts, Seed and Range for Guidance

Styles Tried:

sai-photograpic
None
cinematic-default

Seed: 52794693

Positive Prompts:
young woman with short red hair in graduated bob style teaching math at a blackboard

young woman with short red hair in posh bob style teaching math at a blackboard

young woman with short red hair in victoria beckham posh bob style teaching math at a blackboard

young woman with short red hair stacked  with very short stacked layers in the back   teaching math at a blackboard

young woman with short red hair stacked  with very short stacked layers in the back   teaching math at a blackboard

Negative Prompt:
bangs curls

…. bangs were sometimes screwed up and curls caused artifacts, so these were filtered on.

Guidance range: 5-18.5 ( found to be best)

Additionally, the results were judged by looking at the images for excess artifacts or distortions such as missing fingers, awkward postures and anything else unrealistic beyond the fact that Stable Diffusion is still limited in terms of representing numbers and letters.

 

 

Buddigar

Buddigar: Stable Confusion

Image Generation, Machine Learning, My Thoughts, , , , , , , , , , , , , , , ,

Humor is the weapon of self-control.

Budgierigar
Budgierigar

While playing around with Stable Diffusion, I realized that I must have misspelled “Budgerigar” and was surprised to see the monstrous images that it generated. Perhaps I accidentally typed “Buddigar” or some other variation, resulting in the strange images below. Some of them resemble monsters, while others appear to be stores from a distant land outside of the USA.

These images fall into the uncanny valley of creepiness. For instance, I heard on a Lex Fridman podcast that image generators sometimes struggle with creating realistic facial expressions. It’s fascinating to observe what Stable Diffusion produces when prompted to generate an image of a person with a surprised expression. Although the results can be unsettling and bizarre, they’re also quite amusing, like this one….

OMG, a Buddigar!!!!!
OMG,I spot a Buddigar!!!!!

Buddigars

The first one is my favorite in terms of shock and awe.

Buddigar: It's like Golem's Pet.
Buddigar: It’s like Golem’s Pet.

Buddigar tale by ChatGPT (mostly, just with a bit of prompting and edits)

There is a mischievous creature named Buddigar. It’s small and furry, with sharp claws and bright, beady eyes that seemed to sparkle with mischief. Buddigar loves nothing more than causing trouble, and it has a talent for doing so in the most unexpected ways.

If you can’t find your keys, Buddigar is probably to blame. It has a habit of sneaking into people’s pockets and snatching whatever shiny object catches its eye. It wasn’t that Buddigar has any use for these trinkets – it simply enjoys hoarding them, like a dragon with its treasure.

And if your computer is running slow, you could bet that Buddigar had something to do with it. It has a knack for hiding files, moving icons, and generally wreaking havoc on any device it comes across. Even the most tech-savvy of users are no match for Buddigar’s cunning ways.

But Buddigar’s mischief isn’t limited to material possessions. It has a strange talent for causing physical ailments, too. If you have a pain in your hip, for instance, you could be sure that Buddigar has something to do with it. It’s not that Buddigar means to hurt anyone – it simply enjoys the chaos that its actions cause.

The funny thing is, Buddigar doesn’t try to hide its misdeeds. When caught in the act, it simply grunts and declares its mischievous intent proudly. “Buddigar!” it exclaims, with a mischievous twinkle in its eye. “Buddigar take keys! Buddigar slow computer! Buddigar make hip hurt!” And then it scampers away, leaving behind a trail of chaos and confusion.

Despite its mischievous nature, however, Buddigar is impossible to hate. There is just something endearing about its playful antics, and the way it seemed to revel in the chaos it caused. Even Gollum, the famously twisted and corrupted creature from Lord of the Rings, would have adored Buddigar’s mischievous ways. It is so ugly it’s almost gone full circle to cute in an uncanny way.

And so, Buddigar continues to roam the land, causing trouble wherever it goes. Some curse its name, while others simply laugh and shake their heads in amusement. But love it or hate it, there was no denying that Buddigar is one of a kind – a little creature with  a talent for mischief that it’s unmatched by any other.

 

More Buddigars…

Writing

ChatGPT at it’s finest: writing short helpful code

ChatGPT, Scripts, Trading, , , , , , , , , , , , , ,

ChatGPT at it’s finest: writing short helpful code without having to do fancy work.

The following is a dialog with ChatGPT. I had this script that I would normally have to hand edit all of the ticker  symbols by adding them to the end of the lines of code. I am sure there is some clever way using sed or awk but, I figured why not get ChatGPT to help.

ChatGPT then was also able to generate the Tags for this post,by feeding the post into a prompt and I used Stable Diffusion to generate the featured image using the blog title as the prompt. Sometimes it is good to use ChatGPT to proofread as well but, this post was short and simple enough not to do that.

User

I want to show you an output, in the left column are the ticker symbols for a number of cryptocurrencies, I want you to remember all of them as they will be used in the next question that I will pose…..

Currency  Price      Available  Hold      Avail in USD  Held in USD
AAVE 98.16          0.0 0.0              0.0
ACH 0.020353          0.0 0.0              0.0
ADA 0.3859          0.0 0.0              0.0
ALGO 0.1341          0.0 0.0              0.0
AMP 0.00248          0.0 0.0              0.0
ANKR 0.0244          0.0 0.0              0.0
ATOM 9.283          0.0 0.0              0.0
AVAX 20.66          0.0 0.0              0.0
AXS 6.8          0.0 0.0              0.0
BAL 3.68          0.0 0.0              0.0
BAND 1.417          0.0 0.0              0.0
BAT 0.21159          0.0 0.0              0.0
BCH 223.72          0.0 0.0              0.0
BNT 0.75          0.0 0.0              0.0
BOND 2.2          0.0 0.0              0.0
BTC 37964.39          0.0 0.0              0.0
CGLD 0.53          0.0 0.0              0.0
COMP 51.17          0.0 0.0              0.0
CRV 0.5658          0.0 0.0              0.0
CVC 0.1102          0.0 0.0              0.0
DAI 0.9999          0.0 0.0              0.0
DOGE 0.08033          0.0 0.0              0.0
DOT 5.263          0.0 0.0              0.0
ETC 18.86          0.0 0.0              0.0
ETH 2054.84          0.0 0.0              0.0
FARM 32.73          0.0 0.0              0.0
FET 0.5347          0.0 0.0              0.0
FIL 4.545          0.0 0.0              0.0
GLM 0.2418          0.0 0.0              0.0
GRT 0.1483          0.0 0.0              0.0
ICP 4.605          0.0 0.0              0.0
KNC 0.7065          0.0 0.0              0.0
LINK 14.588          0.0 0.0              0.0
LOOM 0.0673          0.0 0.0              0.0
LPT 6.33          0.0 0.0              0.0
LRC 0.216          0.0 0.0              0.0
LTC 69.78          0.0 0.0              0.0
MANA 0.4405          0.0 0.0              0.0
MATIC 0.7528          0.0 0.0              0.0
MIR 0.0941          0.0 0.0              0.0
MKR 1483.6          0.0 0.0              0.0
MLN 15.12          0.0 0.0              0.0
NKN 0.1064          0.0 0.0              0.0
NMR 15.68          0.0 0.0              0.0
OGN 0.12922          0.0 0.0              0.0
OMG 1.7217          0.0 0.0              0.0
OXT 0.0724          0.0 0.0              0.0
PLA 0.204          0.0 0.0              0.0
POLY 0.1963          0.0 0.0              0.0
QNT 99.56          0.0 0.0              0.0
REN 0.0823          0.0 0.0              0.0
REQ 0.083          0.0 0.0              0.0
RLC 1.4794          0.0 0.0              0.0
RLY 0.0126          0.0 0.0              0.0
SKL 0.0436          0.0 0.0              0.0
SNX 3.392          0.0 0.0              0.0
SOL 58.2          0.0 0.0              0.0
STORJ 0.7658          0.0 0.0              0.0
UMA 1.953          0.0 0.0              0.0
UNI 6.161          0.0 0.0              0.0
USD              0.01 0.0
USDC              0.0 0.0
XLM 0.118572          0.0 0.0              0.0
XTZ 0.817          0.0 0.0              0.0
YFI 8519.94          0.0 0.0              0.0
ZEC 29.03          0.0 0.0              0.0
ZRX 0.39338          0.0 0.0              0.0

I don’t have the ability to remember information from one interaction to the next, but I can certainly help you analyze or work with the data you’ve provided in this session. If you have a specific question or task related to this cryptocurrency data, feel free to let me know, and I’ll do my best to assist you!

User

Can you print out all of the symbols in the left column?

Certainly! Here are all the ticker symbols from the left column:

objectivec
AAVE, ACH, ADA, ALGO, AMP, ANKR, ATOM, AVAX, AXS, BAL, BAND, BAT, BCH, BNT, BOND, BTC, CGLD, COMP, CRV, CVC, DAI, DOGE, DOT, ETC, ETH, FARM, FET, FIL, GLM, GRT, ICP, KNC, LINK, LOOM, LPT, LRC, LTC, MANA, MATIC, MIR, MKR, MLN, NKN, NMR, OGN, OMG, OXT, PLA, POLY, QNT, REN, REQ, RLC, RLY, SKL, SNX, SOL, STORJ, UMA, UNI, USD, USDC, XLM, XTZ, YFI, ZEC, ZRX

If you have any specific tasks or questions related to these symbols, feel free to let me know!

User

OK, good, you remembered them. Now for the next step. I am going to show an example using btc and I would like you to create a listing with all of the ticker symbols from the left column. Example: $DIR/rsi-fractals-avgs-bbands-txt-summary.py btc

Sure, I understand. If you could provide an example of the desired output or clarify the specific format or information you’re looking for related to each ticker symbol (similar to the example you provided with btc), I’d be happy to assist you further.

User

I am looking for you to create code that has $DIR/rsi-fractals-avgs-bbands-txt-summary.py on every line but with the ticker symbols from the left column appear on that line right after it.

Certainly! It sounds like you want a list of commands where $DIR/rsi-fractals-avgs-bbands-txt-summary.py is followed by each ticker symbol. Here’s an example Python script that generates those commands:

python
# List of ticker symbols ticker_symbols = [     "AAVE", "ACH", "ADA", "ALGO", "AMP", "ANKR", "ATOM", "AVAX", "AXS", "BAL",     "BAND", "BAT", "BCH", "BNT", "BOND", "BTC", "CGLD", "COMP", "CRV", "CVC",     "DAI", "DOGE", "DOT", "ETC", "ETH", "FARM", "FET", "FIL", "GLM", "GRT", "ICP",     "KNC", "LINK", "LOOM", "LPT", "LRC", "LTC", "MANA", "MATIC", "MIR", "MKR",     "MLN", "NKN", "NMR", "OGN", "OMG", "OXT", "PLA", "POLY", "QNT", "REN", "REQ",     "RLC", "RLY", "SKL", "SNX", "SOL", "STORJ", "UMA", "UNI", "USD", "USDC", "XLM",     "XTZ", "YFI", "ZEC", "ZRX" ] # Directory variable DIR = "$DIR" # Generate commands commands = [f"{DIR}/rsi-fractals-avgs-bbands-txt-summary.py {symbol}" for symbol in ticker_symbols] # Print commands for command in commands:     
       print(command)

Replace $DIR with the actual directory you want to use. This script will generate commands for each ticker symbol, following the format you provided. If you have any specific adjustments or requirements, feel free to let me know!

OK how did it turn out?

This is the output, just as I expected and was looking for. Saved me probably about 15 minutes of tedious hand editing…

$DIR/rsi-fractals-avgs-bbands-txt-summary.py AAVE
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ACH
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ADA
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ALGO
$DIR/rsi-fractals-avgs-bbands-txt-summary.py AMP
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ANKR
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ATOM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py AVAX
$DIR/rsi-fractals-avgs-bbands-txt-summary.py AXS
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BAL
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BAND
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BAT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BCH
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BNT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BOND
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BTC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py CGLD
$DIR/rsi-fractals-avgs-bbands-txt-summary.py COMP
$DIR/rsi-fractals-avgs-bbands-txt-summary.py CRV
$DIR/rsi-fractals-avgs-bbands-txt-summary.py CVC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py DAI
$DIR/rsi-fractals-avgs-bbands-txt-summary.py DOGE
$DIR/rsi-fractals-avgs-bbands-txt-summary.py DOT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ETC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ETH
$DIR/rsi-fractals-avgs-bbands-txt-summary.py FARM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py FET
$DIR/rsi-fractals-avgs-bbands-txt-summary.py FIL
$DIR/rsi-fractals-avgs-bbands-txt-summary.py GLM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py GRT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ICP
$DIR/rsi-fractals-avgs-bbands-txt-summary.py KNC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LINK
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LOOM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LPT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LRC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LTC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MANA
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MATIC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MIR
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MKR
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MLN
$DIR/rsi-fractals-avgs-bbands-txt-summary.py NKN
$DIR/rsi-fractals-avgs-bbands-txt-summary.py NMR
$DIR/rsi-fractals-avgs-bbands-txt-summary.py OGN
$DIR/rsi-fractals-avgs-bbands-txt-summary.py OMG
$DIR/rsi-fractals-avgs-bbands-txt-summary.py OXT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py PLA
$DIR/rsi-fractals-avgs-bbands-txt-summary.py POLY
$DIR/rsi-fractals-avgs-bbands-txt-summary.py QNT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py REN
$DIR/rsi-fractals-avgs-bbands-txt-summary.py REQ
$DIR/rsi-fractals-avgs-bbands-txt-summary.py RLC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py RLY
$DIR/rsi-fractals-avgs-bbands-txt-summary.py SKL
$DIR/rsi-fractals-avgs-bbands-txt-summary.py SNX
$DIR/rsi-fractals-avgs-bbands-txt-summary.py SOL
$DIR/rsi-fractals-avgs-bbands-txt-summary.py STORJ
$DIR/rsi-fractals-avgs-bbands-txt-summary.py UMA
$DIR/rsi-fractals-avgs-bbands-txt-summary.py UNI
$DIR/rsi-fractals-avgs-bbands-txt-summary.py USD
$DIR/rsi-fractals-avgs-bbands-txt-summary.py USDC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py XLM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py XTZ
$DIR/rsi-fractals-avgs-bbands-txt-summary.py YFI
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ZEC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ZRX

What’s it for?

It is screener code that will go into a script that prints out all the cryptos in the list sorted by their position relative to their 20 day moving average. It uses the same logic as used for the Mayer Multiple, except it is on a 20 day MA basis.

X = Current Price / 20DMA

Code

#!/bin/bash 
DIR="/home/erick/python/ta"


rm $DIR/output-rsi-er-price-sma20-multiple.csv

$DIR/rsi-fractals-avgs-bbands-txt-summary.py AAVE
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ACH
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ADA
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ALGO
$DIR/rsi-fractals-avgs-bbands-txt-summary.py AMP
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ANKR
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ATOM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py AVAX
$DIR/rsi-fractals-avgs-bbands-txt-summary.py AXS
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BAL
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BAND
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BAT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BCH
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BNT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BOND
$DIR/rsi-fractals-avgs-bbands-txt-summary.py BTC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py CGLD
$DIR/rsi-fractals-avgs-bbands-txt-summary.py COMP
$DIR/rsi-fractals-avgs-bbands-txt-summary.py CRV
$DIR/rsi-fractals-avgs-bbands-txt-summary.py CVC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py DAI
$DIR/rsi-fractals-avgs-bbands-txt-summary.py DOGE
$DIR/rsi-fractals-avgs-bbands-txt-summary.py DOT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ETC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ETH
$DIR/rsi-fractals-avgs-bbands-txt-summary.py FARM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py FET
$DIR/rsi-fractals-avgs-bbands-txt-summary.py FIL
$DIR/rsi-fractals-avgs-bbands-txt-summary.py GLM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py GRT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ICP
$DIR/rsi-fractals-avgs-bbands-txt-summary.py KNC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LINK
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LOOM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LPT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LRC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py LTC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MANA
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MATIC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MIR
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MKR
$DIR/rsi-fractals-avgs-bbands-txt-summary.py MLN
$DIR/rsi-fractals-avgs-bbands-txt-summary.py NKN
$DIR/rsi-fractals-avgs-bbands-txt-summary.py NMR
$DIR/rsi-fractals-avgs-bbands-txt-summary.py OGN
$DIR/rsi-fractals-avgs-bbands-txt-summary.py OMG
$DIR/rsi-fractals-avgs-bbands-txt-summary.py OXT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py PLA
$DIR/rsi-fractals-avgs-bbands-txt-summary.py POLY
$DIR/rsi-fractals-avgs-bbands-txt-summary.py QNT
$DIR/rsi-fractals-avgs-bbands-txt-summary.py REN
$DIR/rsi-fractals-avgs-bbands-txt-summary.py REQ
$DIR/rsi-fractals-avgs-bbands-txt-summary.py RLC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py RLY
$DIR/rsi-fractals-avgs-bbands-txt-summary.py SKL
$DIR/rsi-fractals-avgs-bbands-txt-summary.py SNX
$DIR/rsi-fractals-avgs-bbands-txt-summary.py SOL
$DIR/rsi-fractals-avgs-bbands-txt-summary.py STORJ
$DIR/rsi-fractals-avgs-bbands-txt-summary.py UMA
$DIR/rsi-fractals-avgs-bbands-txt-summary.py UNI
$DIR/rsi-fractals-avgs-bbands-txt-summary.py USD
$DIR/rsi-fractals-avgs-bbands-txt-summary.py USDC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py XLM
$DIR/rsi-fractals-avgs-bbands-txt-summary.py XTZ
$DIR/rsi-fractals-avgs-bbands-txt-summary.py YFI
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ZEC
$DIR/rsi-fractals-avgs-bbands-txt-summary.py ZRX


echo "Cur  ,Und,Targt,RSI, ER,20Ml,Mayer Mult."


# Sort by desending price/sma20 multiples
sort -nk6 -t, -r $DIR/output-rsi-er-price-sma20-multiple.csv

the script runs the rsi-fractals-avg-bbands-txt-summary code on all the tickers listed and saves output to…

output-rsi-er-price-sma20-multiple.csv

then sorts the output by the 20sma multiple column

 

Results

Best relative performers on top, worst on bottom. This idea comes from the book Fortune’s Formula and appears as a technique used by Princeton-Newport. Basically sell the top and buy the bottom ( deep value…maybe), or alternativly short the top and long the bottom. Or in the case of crypto the bottom ones might be just be duds that are always behind the pack, maybe don’t even bother with them.

Cur  ,Und,Targt,RSI, ER,20Ml,Mayer Mult.
UNI  ,USD,False,73,0.06,1.18,1.31
SKL  ,USD,False,63,0.07,1.1 ,1.59
FET  ,USD,False,67,0.14,1.09,1.97
SNX  ,USD,False,63,0.09,1.07,1.43
GRT  ,USD,False,60,0.05,1.07,1.37
AXS  ,USD,False,56,0.0 ,1.07,1.2
AMP  ,USD,False,63,0.05,1.06,1.17
MKR  ,USD,False,60,0.13,1.05,1.34
AVAX ,USD,False,61,0.07,1.05,1.61
RLC  ,USD,False,61,0.01,1.04,1.19
OGN  ,USD,False,56,0.05,1.03,1.33
UMA  ,USD,False,52,0.07,1.01,1.13
NMR  ,USD,False,55,0.02,1.01,1.14
MANA ,USD,False,48,0.01,1.01,1.18
DOGE ,USD,False,55,0.02,1.01,1.15
BTC  ,USD,False,54,0.14,1.01,1.27
REQ  ,USD,False,52,0.02,1.0 ,1.1
NKN  ,USD,False,59,0.02,1.0 ,1.1
ICP  ,USD,False,51,0.04,1.0 ,1.14
FARM ,USD,False,53,0.04,1.0 ,1.31
ETH  ,USD,False,47,0.04,1.0 ,1.13
DAI  ,USD,False,44,0.01,1.0 ,1.0
ATOM ,USD,False,49,0.05,1.0 ,1.09
ADA  ,USD,False,57,0.01,1.0 ,1.25
AAVE ,USD,False,52,0.08,1.0 ,1.38
STORJ,USD,False,46,0.13,0.99,2.0
SOL  ,USD,False,51,0.19,0.98,2.12
ALGO ,USD,False,47,0.06,0.98,1.13
ACH  ,USD,False,51,0.02,0.98,1.07
ZEC  ,USD,False,43,0.04,0.97,1.0
XLM  ,USD,False,43,0.05,0.97,1.03
LTC  ,USD,False,42,0.03,0.97,0.89
LINK ,USD,False,46,0.17,0.97,1.8
DOT  ,USD,False,45,0.01,0.97,1.08
COMP ,USD,False,49,0.05,0.97,1.11
QNT  ,USD,False,43,0.02,0.96,0.98
PLA  ,USD,False,45,0.02,0.96,1.19
OXT  ,USD,False,47,0.01,0.96,1.13
ETC  ,USD,False,39,0.0 ,0.96,1.07
CRV  ,USD,False,47,0.08,0.96,0.92
BCH  ,USD,False,36,0.1 ,0.96,1.08
LRC  ,USD,False,42,0.08,0.95,0.99
FIL  ,USD,False,41,0.0 ,0.95,1.14
CGLD ,USD,False,44,0.0 ,0.95,1.07
BAL  ,USD,False,38,0.12,0.95,0.88
XTZ  ,USD,False,38,0.03,0.94,1.03
MLN  ,USD,False,36,0.05,0.94,0.9
BAT  ,USD,False,38,0.01,0.94,1.07
BAND ,USD,False,44,0.01,0.94,1.12
ANKR ,USD,False,43,0.01,0.94,1.06
LPT  ,USD,False,41,0.04,0.92,1.13
GLM  ,USD,False,41,0.02,0.92,1.18
CVC  ,USD,False,48,0.05,0.91,1.27
ZRX  ,USD,False,36,0.09,0.9 ,1.68
YFI  ,USD,False,42,0.04,0.9 ,1.32
MATIC,USD,False,28,0.03,0.9 ,1.1
KNC  ,USD,False,27,0.03,0.9 ,1.11
BNT  ,USD,False,32,0.08,0.9 ,1.56
POLY ,USD,False,30,0.01,0.81,1.2

 

 

ChatGPT will create Tags

Finally by posting this post into a prompt you can ask ChatGPT to generate Tags for the post. I used most of them but trimmed a few that were redundant.

makemore

Makemore

Machine Learning, Text Generation, , , , , , , , , , , , , , , , , , ,

The best way to predict a future is to look for it. We live in the moment where the world is a vast machine for predicting the future.

This summer, I spent a significant amount of time contemplating large language models and delving deeper into their research. My first encounter with GPT-2 was back in 2019, where I explored its code and experimented with it. During this period, I became curious about transfer learning and its applications. Additionally, I had some prior knowledge about transformers, but it wasn’t as comprehensive as my understanding of LSTMs and RNNs. I couldn’t confidently explain what they did, for example.

While researching transfer learning with smaller models like GPT-2, I stumbled upon Gwern Branwen’s website (https://gwern.net/) and, in particular, his TWDNE Project (https://gwern.net/twdne). I found it clever because it combined a generative model for both images and text. I decided to focus on the text side of the project, as the image aspect was already well-addressed by applications like Stable Diffusion….

Misato Katsuragi as a Math Teacher
Misato Katsuragi as a Math Teacher

I might revisit the image style transfer aspect in the future, as I had previously explored it to some extent. You can find more about this in my “How to Generate Art Demo Followup.”

Before this, I had predominantly explored machine learning with code from the ground up using Python (PMLC). I have used ML practically in the form of genetic algorithms for tuning parameters on investing models for years, non-differentiable, so no chain rule! An offshoot was a project called gen-gen-algo, a generic genetic algorithm. Now, finally after all these side quests, I was ready to tackle something more complex and cutting-edge using GPT.

I found excellent resources on GitHub and in video format from Andrej Karpathy (https://github.com/karpathy). The following repositories were particularly helpful in my learning journey. The first one, “nn-zero-to-hero,” features a series of videos that provided a solid foundation in understanding transformers.

The second repository, “makemore,” served as my warm-up exercise to get back into working with transformers and Large Language Models (LLMs) after a period of dormancy in the field. You can access these repositories here:

1. “nn-zero-to-hero”: https://github.com/karpathy/nn-zero-to-hero
2. “makemore”: https://github.com/karpathy/makemore

Fork of makemore

My experience with “makemore” went beyond the basic examples provided in the original repository, which generated new names based on a dataset of names. Initially, my goal was to apply “makemore” to various datasets other than “names.txt.” I experimented with larger and smaller datasets, including those with extensive collections of English words, numbers for addition, square roots, and a substantial dataset of quotes containing nearly 10 million entries, some of which had lines as long as 505 characters. By using scripts and modifications to “makemore.py,” I conducted a grid search to optimize hyperparameters, including constraints on model size. Output from “makemore.py” was saved to a CSV file, along with hexadecimal hash values for easy tracking and analysis during the tuning process.

To further enhance the code, I introduced a grid search optimization method using a Bash script. This allowed for exploring the hyperparameter space while maintaining a ceiling on the model size. Without such constraints, optimization typically led to increasingly larger models that resulted in the lowest loss.

I also introduced the concept of assigning a random hexadecimal tag to the output of “makemore.py.” This tagging system facilitated the easy identification of the best loss and the associated set of hyperparameters that produced it. Additionally, I incorporated an early stopping mechanism into the “makemore.py” code.

If you’re interested in exploring my fork of Andrej Karpathy’s “makemore” code, you can find it here:

https://github.com/erickclasen/makemore

For a more detailed understanding, I’ve created a comprehensive “verbose-readme.pdf” that provides additional information:

Version on this site, opens in browser:

verbose-readme

GitHub Version requires downloading:

https://github.com/erickclasen/makemore/blob/master/verbose-readme.pdf

 

 

 

 

Secure computer on user side

Taking Control of Your Online Experience: The Case for Client-Side Content Filtering

Machine Learning, My Thoughts, , , , , , , , , , , , , ,

I love technology. We love movies, we love television, we love technology.
We are a society where technology is used in an almost inhuman manner.

It appears that social media companies are implementing filters to regulate content, with the intention of creating a more suitable experience for consumers and curbing the dissemination of misinformation. However, the COVID-19 pandemic revealed a potential for bias, suppression, and censorship in this process. Such manipulation could also extend to political motives, where information might be skewed to suit a particular agenda.

Control should be on the users side

This raises the question: Why employ content filtering on the server side when it could be done on the client side? Could users have settings that allow them to customize content filtering based on their preferences? There are valid reasons for such an approach. Users may wish to filter out redundant, repetitive, or annoying content, similar to how ad blockers or spam filters function, but on a more sophisticated scale.

This concept could involve a machine learning component, perhaps utilizing a Bayesian filter. Users could prime it with specific keywords, examples of unwanted posts, and other content to be avoided. This filter would reside within the user’s computer, phone, or even as a browser plug-in. By processing the content at the browser level, the filter could operate effectively, unlike at the router level where encrypted content would be inaccessible. This client-side filter would work similarly to a firewall, analyzing and potentially blocking or redirecting content based on predefined criteria.

AI in the users toolbox

Considering concerns about AI alignment and safety, it’s important to note that the genie is already out of the bottle. Machine learning resources and knowledgeable individuals are widespread. Thus, rather than trying to rein in AI development, the focus should be on leveraging it to the user’s advantage. While organizations like OpenAI strive for responsible AI, and governments aim to establish their standards, consumers require a means to navigate through the information landscape. Users would also want to choose their own alignment rather than have a government or corporation choose it for them as a form of soft censorship.

Fake out the Fake News

This approach could even encompass traditional media, which sometimes appears to convey biased narratives. To address this, a client-side filter could be developed. The name for such a tool is open to discussion—whether it’s dubbed a firewall, content blocker, or a “BS Defender.” Regardless of its name, the need for a user-configurable, trainable, and adaptive filter is evident. This filter could operate through reinforcement learning via human preference selection and incorporate Bayesian adaptive learning.

Next Steps

For someone with coding experience spanning four decades, the framework for such a tool is already something that I can imagine on a high level and it could be done in a few different ways. Right now, I am still thinking on it. The potential exists for this idea to evolve into a comprehensive white paper. On the coding side, a proof of concept could be created, perhaps using Python, to showcase the core filtering concept in action. Sharing this concept for consideration and exploration is crucial in a time when content filtering, covering everything from traditional media to social media and ads, is becoming increasingly important. An adaptable and user-driven filter, bridging the gap between the browser and user experience, holds immense value.

Wave Styled by ML

Modifying a image with a style using machine learning

Machine Learning, , , , , , , , , , ,

The first ingredient to happiness in happiness is an attitude of openness.

Fun With Machine Learning

I have been sharing the following images around. They were created with machine learning code that I came across and modified a bit while examining how it operates.

See the following for more info… How to Generate Art Demo Command Line Version

The code basically takes an image and imposes a style from another image upon it. It is rather computational expensive as it takes around 12 Gigabytes of RAM to “work” on a 1024×1024 pixel image and about 2 hours of compute time to run the 20 iterations required to complete an image. Machine learning is a fascinating new frontier in technology that I have been spending some time since the spring of 2018 getting to understand on a deeper level. I’ve seen a lot of technologies come and go but, this field has stunned me, is moving forward very fast and is here to stay.

Prior to 2018 I had some exposure to machine learning in the sense of using adaptive control systems in industry. I also worked on a research project that involved a type of fuzzy logic and cellular automata for a learning system that would be used in a control loop. I also developed code that used a tractrix curve as the main element of a control system. But, that is kind of simple machine learning as compared to what is going on today.

This link is to the original paper on this topic, there are some more images in it as well,

https://arxiv.org/pdf/1508.06576.pdf?fbclid=IwAR19_eqx3SmHMyqnAnirbfyAWDqLYkqMp98C0XZ6GEfeXZRxnDPaagZ5B_E

RNN Text Generation Using Tensorflow

Machine Learning, Text Generation, , , ,

Imagination is the power to make a difference in yourself.

After trying a few RNNa and LSTMs for text generation that rely on numpy alone it is interesting to see the performance of Tensorflow based code that is closer to the cutting edge of what is possible to do with machine learning.

I found a good and easy to use set of code in the following Github archive…

https://github.com/spiglerg/RNN_Text_Generation_Tensorflow

The requirements are simple…

numpy==1.13.3
tensorflow==1.4.0

I was running it on Conda Python 3.6 environment but, this is not a requirement. The code uses a saved folder where it can save training checkpoints, so it is possible to interrupt and resume training and also use it in a generate or “talk” mode after the model has been trained. The caveat that I learned quickly when training on a few types of files is when training each type of file that is trained into requires it’s own set of checkpoints, which is pretty obvious. So it is best to either wipe out the saved dir contents after a run on a specific corpus. OR, better yet make a subdir for the training checkpoints.

Training is basically sending it the following command…

python rnn_tf.py --input_file=data/us-constitution.txt --ckpt_file="saved/model.ckpt" --mode=train

Once trained it can be fed another command…

python rnn_tf.py --input_file=us-const-lstm/us-constitution.txt --ckpt_file="saved/model.ckpt" --test_prefix="The " --mode=talk

or if the checkpoint files have been moved to their own directory then you can use something like this…

python rnn_tf.py --input_file=us-const-lstm/us-constitution.txt --ckpt_file="saved/us-const-trained/model.ckpt" --test_prefix="The " --mode=talk

In the structure for the commands the location of the file is listed and the location of the checkpoint file as well. The generate mode allows priming with a word or phrase such as “The”.

The US Constitution is not a big corpus and I am sure this code like others would benefit from training against a larger corpus. My intent in the future for an experiment is to train it against a file containing all the posts on this site to see what it can do on that corpus.

-rw-r–r– 1 erick erick 1115394 Apr 26 12:01 shakespeare.txt
-rw-r–r– 1 erick erick   45120 Apr 26 12:59 us-constitution.txt
-rw-r–r– 1 erick erick  374605 Apr 26 13:52 my-posts.txt

When trained on the US Constitution it does very well at producing coherent text. Besides the lack of capitalization it seems to be actually to the point of memorizing parts of the text. This might be because it is a small corpus and it is overfitting.

The Senators and Representatives before mentioned, and the Members of the
several State Legislatures, and all executive and judicial Officers, both of
the United States and of the several States, shall be bound by Oath or
Affirmation, to support this Constitution; but no religious Test shall ever be
required as a Qualification to any Office or public Trust under the United
States.

Article 7.

The Ratification of the Conventions of nine States, shall be sufficient for the
Establishment of this Constitution between the States so ratifying the Same.

Sentence:
the several states, shall be bound by oath or
affirmation, to support this constitution; but no religious test shall ever be
required as a qualification to any office or public trust under the united
states.

article 7.

the ratification of the conventions of nine states, shall be sufficient for the
establishment of this constitution between the states so ratifying the same.

done in convention by the unanimous consent of the states present the
seventeenth day of september in the year of our lord on

 

 

the Case of a Bill.

Section 8
The Congress shall have Power To lay and collect Taxes, Duties, Imposts and
Excises, to pay the Debts and provide for the common Defence and general
Welfare of the United States; but all Duties, Imposts and Excises shall be
uniform throughout the United States;

To borrow money on the credit of the United States;

To regulate Commerce with foreign Nations, and among the several States, and
with the Indian Tribes;

To establish an uniform Rule of Naturalization, and un

Sentence:
the case of a bill.

section 8
the congress shall have power to lay and collect taxes, duties, imposts and
excises, to pay the debts and provide for the common defence and general
welfare of the united states; but all duties, imposts and excises shall be
uniform throughout the united states;

to borrow money on the credit of the united states;

to regulate commerce with foreign nations, and among the several states, and
with the indian tribes;

to establish an uniform rule of naturalization, and un

 

Training

Training against the corpus of blog posts on this site produced output like this and took about 4 hours of compute time.

batch: 0  loss: 4.492201328277588  speed: 121.8853488969507 batches / s
batch: 100  loss: 3.214789628982544  speed: 1.3747759497226923 batches / s
batch: 200  loss: 3.0983948707580566  speed: 1.4065962415903654 batches / s
batch: 300  loss: 2.8669371604919434  speed: 1.4141226357348917 batches / s
batch: 400  loss: 2.359729051589966  speed: 1.416853411853437 batches / s
batch: 500  loss: 2.0080957412719727  speed: 1.4160802277642834 batches / s

…

batch: 19500  loss: 0.22069120407104492  speed: 1.4188681716674931 batches / s
batch: 19600  loss: 0.21757778525352478  speed: 1.4218841226396346 batches / s
batch: 19700  loss: 0.2309599369764328  speed: 1.362554971973392 batches / s
batch: 19800  loss: 0.23969298601150513  speed: 1.3983937654375616 batches / s
batch: 19900  loss: 0.23989509046077728  speed: 1.3854887855619515 batches / s

The following is some samples of the output it generates. It definitely could use more training to help it. The fact that the posts contain some code, numbers and jargon probably doesn’t help either.

Sentence:
the installed.
 display install wiflinut for ray run process queue every monday, wednesday and friday ran
   1000000 ractine resitely and configure a firewall to only allow certain ip numbers a
   connection to show that the board is
   powered. there are a concatenated version of the log.txt
cacking out of the full -ho 1 than i could have it may be set the
   command which just restart the “how ther have up suncals
   regulator, frequency valies.
   more data. i sho, vift…
sudo selond

   below is

Sentence:
the whole hmad can noid through the server and logged
in a while later and the shutdown script had recorded failed pings into
systemctl.

i was not ne rewent when it shuts down.

for a help afout shourd entire (but mean most looking a series of
for clean ubuntu server install will prompt for a username and password to access folders as
well, especially if the users and password is needed autosuspend should oright. it level, no 62 defanly 34-fermentation crontab, still radio
shar

 

 

 

Forest Gallery

How to Generate Art Demo Followup

Image Generation, Machine Learning, , , ,

Only the imagination grows out of its limitations.

In the example shown in the previous post I used 20 iterations at 512×512. A few lingering questions that might be asked are…

  1. What about more iterations?
  2. What about a lower resolution, like 256×256 ?
  3. Machine learning code typical initializes using random parameters will this affect the image in another identical run ?

 

20 Iterations and 50 Iterations

More iterations up to a point make for a better image. There is a point where the loss value deltas get smaller between iterations and a point of diminishing returns is reached. Not much difference that can be seen happens beyond 20 iterations for this run. There are minor details that have changed but you have to really look carefully to pick them out.

Start of iteration 0
Current loss value: 1.68853e+11
Iteration 0 completed in 672s
Start of iteration 1
Current loss value: 1.06826e+11
Iteration 1 completed in 616s
Start of iteration 2
Current loss value: 7.61243e+10
Iteration 2 completed in 594s
Start of iteration 3
Current loss value: 5.69757e+10
Iteration 3 completed in 501s
Start of iteration 4
Current loss value: 4.73256e+10
Iteration 4 completed in 496s
Start of iteration 5
…..
Start of iteration 9
Current loss value: 3.22461e+10
Iteration 9 completed in 498s

……
Start of iteration 19
Current loss value: 2.63259e+10
Iteration 19 completed in 471s
…….
Start of iteration 49
Current loss value: 2.26513e+10
Iteration 49 completed in 592s

Lower Resolution

The model will perform poorly on lower resolutions, even with 20 iterations, 256×256 will look sloppy and abstract. The only reason to go this low would be to run a bunch of iterations fast to see if it worth trying at higher resolution. Kind of like a preview. On my machine the 256×256 iterations run about 5x faster than the 512×512 iterations.

Random Initialization

Because the machine learning model loads itself with random weights and biases at the start of a model run and works from that as a starting point there is some variations in the results from run to run. This can be seen in theses images as there are slight variations in the results. Sometimes it is worth running the model over and over and then hand picking the best result from a batch of outputs.

As an aside. In some machine learning code it is possible to seed the random number generator so that the random starting point is not really random but seeded to be able to reproduce the same results. Occasionally I have had to do this when training genetic algorithms for trading, hand built code, so total control on my part. I basically want a reference run against which I can gauge future changes to the code against. By using a standard set of input, a fixed time series and seeded random initialization, I get the same tuning every time. Then if a change happens in the code, I know it is a code change and not in the data. Have a reference copy archive makes it reproducible.

Stacks of Coin

Bitcoin’s Soft Underbelly

ChatGPT, Cryptocurrency, My Thoughts, , , ,

The best way to make money is to go out and do it yourself.

Hey ChatGPT, Do you what Bitcoin is?

Yes, I am familiar with Bitcoin. Bitcoin is a decentralized digital currency that uses cryptography for security and is supported by a peer-to-peer network. It was created in 2009 by an unknown individual or group of individuals using the pseudonym Satoshi Nakamoto.

Bitcoin is based on a distributed ledger technology called the blockchain, which allows it to operate without the need for a central authority. Transactions are recorded in blocks, which are linked together in a chain. This chain of blocks is secured using complex mathematical algorithms, which makes it extremely difficult to alter the record of transactions.

Bitcoin can be used to send and receive payments over the internet, and it has attracted a lot of attention due to its potential to disrupt traditional financial systems. However, it has also been associated with criminal activities, and its value can be volatile.

Hidden Risks

Just to linger on the topic for a moment, Bitcoin, has another risk that does not get enough mention. Cryptography has a shelf life and eventually breaks, either via some discovery of a short path to crack a private key given a public key or collisions. Bitcoin uses SHA256 encryption which seems safe for now but, as with all hash algorithms, there is no closed form solution that shows it is watertight. It’s more along the lines of, well the last cryptographic code broke because of X,Y or Z, so lets beef the new one up to be stronger in those areas. But, someone might find a crack in SHA256 tomorrow, just like what happened for SHA-1 not too long ago From what I recall it was a university research project that broke it. Or looking at the second failure mode, when will brute force collision with a large number of private keys become possible and profitable, It’s only a matter of time. 256 bits is a lot a space to brute force through. But, computing power (Now multi threaded quantum computing too ) grows over time and large entities such as nation states with deep pockets could do something like this at some point in the not so distant future. Is this possible, who knows now, but someday it might.

There are known knowns, known unknowns and then the ones that get you when you least expect it, unknown unknowns and this risk clearly falls in that category.

Interesting to Skim

https://www.whitehouse.gov/wp-content/uploads/2022/11/M-23-02-M-Memo-on-Migrating-to-Post-Quantum-Cryptography.pdf

wave-1

Wave

Image Generation, Machine Learning, , , ,

wave-galleryOnly the imagination grows out of its limitations.

A while ago I played around with some code that was based on the following paper. One of the first things that I tried it on was a sample of the famous wave painting, styled by an actual wave.

A Neural Algorithm of Artistic Style

Leon A. Gatys,1,2,3∗Alexander S. Ecker,1,2,4,5Matthias Bethge1,2,41Werner Reichardt Centre for Integrative Neuroscienceand Institute of Theoretical Physics, University of T ̈ubingen, Germany2Bernstein Center for Computational Neuroscience, T ̈ubingen, Germany3Graduate School for Neural Information Processing, T ̈ubingen, Germany4Max Planck Institute for Biological Cybernetics, T ̈ubingen, Germany5Department of Neuroscience, Baylor College of Medicine, Houston, TX, USA∗

Fine art, especially painting, humans have mastered the skill to create unique visual experiences through composing a complex interplay between the con-tent and style of an image. Thus far the algorithmic basis of this process is unknown and there exists no artificial system with similar capabilities. How-ever, in other key areas of visual perception such as object and face recognition near-human performance was recently demonstrated by a class of biologically inspired vision models called Deep Neural Networks.1, 2Here we introduce an artificial system based on a Deep Neural Network that creates artistic images of high perceptual quality. The system uses neural representations to sepa-rate and recombine content and style of arbitrary images, providing a neural algorithm for the creation of artistic images. Moreover, in light of the strik-ing similarities between performance-optimised artificial neural networks and biological vision,3–7our work offers a path forward to an algorithmic under-standing of how humans create and perceive artistic imagery.1arXiv:1508.06576v2 [cs.CV] 2 Sep 2015

https://arxiv.org/pdf/1508.06576.pdf

Animation of wave image being styled
Animation of wave image being styled
Frames from the iterative process of creating the resultant image

See How to Generate Art Demo Command Line Version for more information.