I'm kind of curious, do you have a fork of the project, since the version I have doesn't support the options -name, -types, etc. I've also been wondering what people have found to be the best training parameters.
No, not a fork. What I sampled from was just a private version of mine where I've tinkered with incorporating neural stacks into the architecture. The main repo can be found here (and encoding repo here), and the sampling script "sample_hs_v3.lua" supports field-specific priming like what I've shown you.
Well for those of you who wanting to do a cube draft, I've still been working a lot at setting mine up; I am currently up to 227 cards, not including lands/artifacts.
I've been busy with a bunch of side projects but I'm trying to set this all up in time for my bachelor party. I'm trying to get an updated version of the corpus encoded in hardcast6drop's format, but for the life of me I can't seem to get it to output the file (my unfamiliarity with linux commands makes this tedious as all hell). If anybody has the encoded corpus that includes up to the new Innistrad set please let me know. Aside from that I do not have a complete collection of the cards posted anywhere but I've included a snapshot of my excel document I'm using to keep track of card count.
EDIT: I am also HIGHLY SUSPICIOUS that some of the Blue and Black cards from this thread were tweaked by the authors and were not simply the raw output from the Neural Net. Like, there was a Blue Serpent creature that was almost word-for-word perfect as a generic needs-islands-to-attack serpent. I tried to exclude those.
I guarantee you that that is evidence of the opposite. If you have a phrase in the corpus that is almost always used in a particular context or in a particular way, the network can overfit and assume that it must always be that way, and will memorize the words verbatim. For example, if I try...
|serpenty serpent||creature||serpent|O|&^^^^/&^^^^|{^^UU^^}|@ can't attack unless defending player controls an island.\{^^UU}: tap target creature and gain control of it until end of turn. that creature gains haste until end of turn.|
|serpenty serpent||creature||serpent|O|&^^^^/&^^^^|{^^UU^^}|@ can't attack unless defending player controls an island.\{^UU}: @ can't be blocked this turn.|
|serpenty serpent||creature||serpent|O|&^^^^/&^^^^|{^UU^^^}|@ can't attack unless defending player controls an island.\{UU}, sacrifice an island: put a &^/&^ green saproling creature token onto the battlefield.|
|serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^^UU^^}|@ can't attack unless defending player controls an island.\when you control no islands, sacrifice @.|
|serpenty serpent||creature||serpent|O|&^^^/&^^^|{^UU^^^}|flying\{^^^}, return a land you control to its owner's hand: uncast target spell unless its controller pays {^} for each card in your hand.|
|serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^UU^^^}|@ can't attack unless defending player controls an island.\{^UU}: @ can't be blocked this turn.|
|serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^UU^^^}|@ can't attack unless defending player controls an island.\{^^UU}: tap target creature an opponent controls.|
|serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^UU^^^}|@ can't attack unless defending player controls an island.\{^^^^^^}, T: search your library for a basic land card, reveal it, and put it into your hand. then shuffle your library.|
|serpenty serpent||creature||serpent|O|&^^^^/&^^^^|{^^UU^^^}|morph {^^^^^^^}|
|serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^UU^^^}|@ can't attack unless defending player controls an island.\when you control no islands, sacrifice @.\{WW}: @ gains flying until end of turn.|
|serpenty serpent||creature||serpent|O|&^^^/&^^^|{^^UU^^}|flying\{^^^}, return a land you control to its owner's hand: uncast target spell unless its controller pays {^} for each card in your hand.|
It happens for the same reason that dragon cards usually have flying. It's not that the network knows that dragons are winged creatures and it would make sense for them to fly, it's that the characters "f l y i n g" in the body text often follow the characters "d r a g o n" in the subtype line.
I assure you that, as a rule, I don't touch the cards. Infecting the results with my own ideas would defeat the purpose of the whole experiment, lol. For the same reason, I'm still working out the graphics thing, so I can get art for cards without having to assign art myself. Ran into a small bug earlier in the week with the model, something's not getting loaded properly when I try and restore the model for sampling. I think it's an issue of my own making though. I'll get it resolved.
I should also say that I'm really excited about how far the Hunter project has come along.
EDIT: If you wanted a Turing test for authenticity of cards, suspicious for me would be a card that is significantly longer than about 160 characters or so, because the network hates run-ons. For the same reason, it can't generate complete planeswalkers or sensible flip cards. That and it's unlikely (though not impossible) for the network to produce cards with multiple interconnected or highly related abilities (aside from "add a charge counter"/"remove a charge counter", etc.). For example, Brimaz, King of Oreskos is unlikely to be produced (unless it's overfitting), simply because the network's attention span and its limited comprehension of the semantics of cards pose significant challenges when it comes to sustained, meaningful creation.
Oh god, the art for Dunelog practically had me in tears. And the art for Ancient Donger is perfect. The bells help it all make sense, lol.
Speaking of art, I'm rerunning the training process to condition the generator on the card vectors. Well, part of the card vectors. I left off the body text because that just muddies the waters too much (there's a correlation between "flying" in the bodytext and the artwork, but not so much "enters the battlefield tapped", etc.). I have to make sure I did everything right, do some tests, etc. But I'll let y'all know how it turns out.
If you're going to do that, shoot me a line, too. I will note, though, that I didn't really see a huge disparity in the number of creatures versus spells; at least not that you wouldn't see in a regular Magic set. Removal tends to be really rare, though, and when it does exist, it's AMAZING. (Coral Witch ftw!) Also, the network loves to make Elves, Goblins, Beasts, Humans, and Horrors, but not so much other tribes. Odd.
Yeah, I can look into that later and let y'all know. And yeah, the distributions that the machine learns and the actual distributions can vary. I mean, within cards, it does a good job, like Dragons and Angels should almost always have flying, but in terms of the distribution of card types, colors, etc. there can be some deviations.
Like you said, I think that some creature types get subsumed into others. Like merfolk and vedalken are functionally similar, but there are more merfolk than vedalken, so if it wants to create a Caller of Gales-esque utility creature, it might be biased towards merfolk even when a vedalken would have worked just as well. Upping the temperature can fix that, but it also causes it to make more mistakes as a result of its frantic excitement.
-------
As an aside, I just got done with fielding questions over Skype reagrding population projections using neural networks for the benefit of folks at a US Census Bureau conference. To paraphrase myself: "I know that our techniques must seem like witchcraft to you... and you would be right! It *is* witchcraft. And it's incredibly effective."
More results to come from me before too long. I've had a slew of personal and professional crises to deal with (these things always seem to happen concurrently). I'm keen on getting things to a point of maturity and stability where we can reliably generate relevant artwork for RoboRosewater cards.
On a related note, I retooled my scripts so that I could feed in animations like GIF files.
Still no face on the Lampshade Lady, but as I pointed out before, I messed up the bounding boxes when we trained the last network, and faces got cut out so often that the network seems to strategically place objects in the way of faces in order to avoid having to draw them. That'll get fixed in the next iteration, and we'll also have increased resolution.
Again, the idea here isn't to try and perfectly reconstruct the input, but to use the input as a means of situating ourselves inside the vector space in a region that structurally/semantically resembles the input. The waterfall, for example, isn't the original waterfall, but happens to be another, similar waterfall inside our miniature MTG-art-inspired universe. You can think of these animations as if you were looking through a tiny pinhole into this bizarre little world.
It's fun to play with, because we never trained the system on animations, and it's nice to see that the concept of motion carries over well. That and it's fun to set people, places, and things on fire.
More to come in the near future. Sorry for the wait.
Oh, and I saw a fun paper by Li Fei-Fei's people at Stanford University on real-time neural style transfer and super-resolution. Like, a speed-up by a factor of 1000. Like I said several months ago, these improvements would come along :D. At this point, I'm just waiting for someone to come out with a Google Cardboard app so I can strap it on and see everything as a Van Gogh painting.
@nyrt: Fascinating find! Thank you for sharing!
@Elseleth: I'm excited to see your robo-cube come to fruition. I can try and pull out all the good non-creature cards that I can find later for you.
Just curious about your approach to increasing the resolution, since you mentioned that it was hard-coded in the network: did you look at all into making the convolutional/deconvolutional part of the structure recursive? I had good luck with adapting the DCGAN code to use recursive deconvolution, which is nice for passing in how many iterations of deconvolution I want without having to retrain the network.
I could have / should have done that. As written, the code has a 32x32 network configuration and a 64x64 network configuration. I changed it so that I could create an NxN configuration for some value of N that I passed. Easy enough. But yes, recursive deconvolution is a thing, and I should look into it. My first goal right now is to condition the network using tags for cards, so that it learns to relate goblin/sorcery/green card artworks with each other. That way, I can control the output by asking it to make something extra "goblin-y", for example. That increases the odds that the art will make sense in the context that I want to use it. I just about have that ready to go; I'm just waiting on the opportunity to run the code again, and that should be soon. Then I can see about restructuring things to get higher resolution images out.
Not me, that's for sure. I recognize some of the names though, like Grefenstette. It looks like the Google Deepmind folks found a way to improve upon our work. I had recommended the use of a bidirectional LSTM, but I've been too busy to get around to testing that. The use of code compression is novel, however. I'm happy that I was able to attract people with adequate funding/time to investigate this topic in more detail.
I'm gonna have to take a look at their implementation whenever that becomes available. I might e-mail them about that later. The numbers look really good.
Thank you for sharing this! This is most helpful.
EDIT(1): I sent an e-mail to the lead authors with my congratulations and inquiries.
EDIT(2): I had a lovely conversation with the folks at DeepMind. Their work has a lot of fun tricks that I could co-opt for our purposes. They're not sure about a timeline for releasing their source code though, as there's a bureaucratic process for all that. I'd have to re-implement it. Which is fine.
EDIT(3): By the way, when I get the art renders working at higher resolutions (and one way or another, I will eventually), I'm going to have so much fun with it. I've found that, when I choose to guide the art generation process with an example, if it doesn't recognize an object in the scene, it tends to replace it with something else that conforms to the geometry. But it's not an exact match, it has artistic freedom (e.g. adding in feet, tail feathers, a mouth, and a dot for an eye). I think this process closely resembles what surrealist André Breton called "pure psychic automatism".
Oh, and I also attached a sample output that I made for one Thraximundar_ of reddit. The art was created by interpolation/inspiration from the art of semantically similar cards. Bit hard to make out, but like I said, that's one of the things I'm working on. The flavor text I kinda cheated on simply because I produced ten of them and hand-picked the best one; I'll need to work out an automated system for that so I don't contaminate RoboRosewater's cards with my ideas, lol.
I'm interested in modifying this to help me generate random monsters for D&D (5e), or at least generate new special abilities, as I'm already able to generate very good base stat blocks. Is the code locked into the given json format and fields, or could it be tweaked for any input/output? If so, where would the changes need to be made?
Took me longer to reply than I had intended, but hello and welcome! I had a look at your source code. I'm most impressed! I admire very deep and nested procedural generations; I can tell there's a lot of attention paid to detail.
I concur with everything that our resident expert (he's modest) maplesmall has told you. There are challenges with using character-level generative models, and it's not always the best fit for every situation. At the very least, I'm not sure that I'd recommend it as an end-to-end solution for you. However, there are plenty of ways that you could incorporate ML tools and techniques into your generation process (piece-wise).
For example, a lot of the stats can be drawn from learned distributions if needed. That being said, you've spent a lot of time and energy calibrating your stat-purchasing model, and I see no reason to throw that out if it works well for you.
Abilities could be generated in a way similar to what we do, though I might recommend a word-level model rather than our character-level one. Or perhaps even some kind of clause-level model, so it'd be like generating an abstract syntax tree of sorts... but I'm not know whether there are any good implementations of that sort available online.
Now, description text, there's something that you're going to be hard-pressed to do in a purely procedural way (unless you want it to come off as very artificial sounding). A generative architecture like ours could churn out monster description text just fine (and you have plenty of data for that), but it wouldn't deliver what you wanted because what it generated would not be conditioned upon or keyed to the monsters, so you'd get fascinating but arbitrary text attached to your monsters. Instead, you'd want to use a conditional neural language models, like the one used here. But instead of picture in -> text out, it'd be monster in -> text out. Same sort of idea. Of course, the technology for that is still maturing, so the results will probably be something of a mixed bag.
If you'd like, you can PM me or send me an email at rmmilewi (at) gmail (dot) com and we can talk more about this in detail. I can see about directing you towards the resources that you'd need.
By the way, while we're on the subject, Alec Radford just put out a lecture entitled "Deep Advances in Generative Modeling". It's a 40 minute presentation that covers virtually all the algorithms that we've been talking about in this thread and then some. If you skip ahead to 34:50, you can see some unpublished results where they condition an image generator on text.
-----
As for me, I've just about got the art-generation-conditioned-on-card-vectors thing coded, but I'm having to wait to run it. Right now I'm training a bunch of LSTM networks to do population forecasting for my state on behalf of some economists and census folks. But once that's done with, I'll be sure to train a new image generation model. Once I get that working, I can see about writing some scripts that'll integrate the card and art generation processes.
EDIT: On a related note, I just got some new hardware in, so that may mean that I'll be able to run Magic experiments in parallel with others. That'll speed things up, lol.
Interesting. I wonder what ongoing training at varied dropout values would do.
One way to find out I guess. (or two if you consider asking someone, but where's the fun in that?)
Another thing to test is to vary the way the representation is organized. The networks I train use a fixed order for all the fields like these:
|undead lior||creature||homunculus|N|&^^/&^|{GU^}|{^^UU}, T: put target creature on top of its owner's library.|
|orochi agent||creature||snake archer|N|&^^/&^^^|{^^GG^}|reach \flash\{WW}: put a +&^/+&^ counter on @.|
|fleetfire elemental||creature||elemental|O|&^^^^/&^|{^^RR^}|haste\{^BB^}: @ gets +&^^/+&^^ until end of turn.|
|master of the master||creature||human knight|A|&^^/&^^|{^WW^^}|first strike, protection from white and from black\as long as @ is on the stack, each creature has "{^WW}, T: @ deals &^ damage to target creature or player."|
You can control how the input is organized by passing different parameters to the mtgencode program.
By the way, Master of the Master is... a very unusual card.
Master of the Master 3W
Creature - Human Knight (Rare)
First strike, protection from white and from black.
As long as Master of the Master is on the stack, each creature has "1W, T: This creature deals 1 damage to target creature or player."
2/2
Are there any cards that resemble it?
Anyway, on the subject of dropout, note that overfitting (where the network copies the cards it studies too closely) can come in many forms. For example, this is what I call a pseudo-clone:
Mesa Enchantress GW
Creature - Human Druid (Rare)
Whenever an enchantment is put into a graveyard from the battlefield, you may put a +1/+1 counter on Mesa Enchantress.
1/2
It has Mesa Enchantress's name and Femeref Enchantress's cost, body, and trigger clause. Just something to look out for when you are comparing results taken from networks trained under different conditions.
When I run training, does it straight afresh, or build off of what it's already learnt?
It can do either. You can either use an existing checkpoint as a starting point, or you can start all over. To start training from an existing checkpoint, use the "-init_from" parameter followed by the path to the checkpoint that you want to use as a starting point.
|5creature|4|6human berserker|7|8&^^/&^|9{WW}, sacrifice @: target creature gets -&^/-&^ until end of turn.\{^^^RR}: put a % counter on @.\countertype % fide|3{^^RR}|0O|1gorgon camph|
The % represents a counter type, which is defined elsewhere in the card. In the case of the card you generated, it should be read as...
Gorgon Camph 2R
Creature - Human Berserker (Common) W, Sacrifice Gorgon Camph: target creature gets -1/-1 until end of turn. 3R: Put a Fide counter on Gorgon Camph.
2/1
Hit my first checkpoint!
...
And another problem.
/home/user/torch/install/bin/luajot: cannot open <cv/custom_format-256//lm_lstm_epoch12.61_0.3874.tz> in mode w at /home/user/torch/pkg/torch/lib/TH/THDiskFile.c:640
Whatever that means.
stack traceback:
{C}: at 0x7f48b4c00e80
{C}: in function 'DiskFile'
/home/user/torch/install/share/lua/5.1/torch/File.lua:385: in function 'save'
train.lua:363: in main chunk
{C}: in function 'dofile'
...aker/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk
{C}: at 0x00405d70
Do those directories exist? The cv subdirectory, followed by the custom_format-256 directory? If not, you could create them with the mkdir command. That might be why you are having an issue.
---
Earlier in the week, I said I'd release a Magic-art-image-generation model and script for playing with it. Then I got hit by an avalanche of work. On the bright side, it's Spring break next week, so that frees up my schedule somewhat. I'm going to look into retraining the system to run with tags (images are passed to the network along with tags like "red" or "goblin", so that way the tags give us a more gentle way of controlling the output image rather than operating on the latent space directly.
Oh, and, as usual, the techinques that I've been using are already obsolete (probably). I just read a fascinating paper by Wang and Gupta entitled "Generative Image Modeling using Style and Structure Adversarial Networks". The novel aspect of their work is that they split up the generator network into two parts. The first is a structure generator that creates the geometry of the image (surface normals). Then that geometry is passed to the style generator that handles textures, lighting, appearance, etc.
Right now, with the setup we've got, style and structure are intermixed. That's why when I change the lighting of a scene, the geometry warps and shifts. Wang and Gupta are able to control these two aspects independently.
For the training data, you'd have to have surface normals for the artwork, but that's actually quite feasible. There are trained models out there for predicting the geometry of 2D images (example). Those systems are usually trained on photographs, but I suspect that they could work decently on artwork. Murk Dwellers no, Rafiq of the Many yes.
Not that I'll mess with that at the moment, but it's nice to know that these options exist.
EDIT: Oh, almost forgot! A team from Google made yet another interesting breakthrough in a paper entitled "One-Shot Generalization in Deep Generative Models". One-shot learning is where you have one or two images of a thing and you then have to arrive at a understanding of what that thing is (or in this case, be able to generate new and interesting versions of it). It's sort of a "holy grail" in computer vision. They're one step closer to achieving that. I've attached an example where the system is presented with a strange alphabet that it has never been trained on, and has to learn to write letters in that alphabet from just those single examples. They did another similar test with faces. Just a few faces, make new faces, that sort of thing. And it does decently, actually. Not bad. There's a lot more that needs to be done in this area, but the pay-off could be great. An example application would be to show a picture of an apple to a robot that has never seen apples before and say "This is an apple. Go find me more of these things. Bring them back to me. Kthxbai."
Forgive my ignorance, but when I follow the install tutorial I get to this point: "luarocks install nngraph"
And hit this snag: "Your user does not have write permissions in[...]"
Halp?
EDIT: I tried "sudo chmod -R ugo+rw /usr/local" and now I just receive "Error no results matching query" when I try "luarocks install nngraph"
EDIT 2:
Went through this workaround, no idea what it does, just found it on google (after much, much searching):
Strange. Well, at least you found a workaround. I'm not sure why it wasn't working for you. Well, the permissions problem I understand, but not the "no results matching query problem."
and replace the X's with the numbers in the name of a checkpoint. But what numbers are those? The one after epoch? But there's only one number after epoch and two sets of X's. :S
The numbers themselves aren't especially important, it's just that you need to feed in a checkpoint file, preferably the most recent one, whichever that is. The numbers at the end should tell you the epoch.
What would decreasing the dropout to 45% do? Just make it more like the existing ones, or totally like the existing ones? And how would one do this?
Dropout is a regularization technique whereby we randomly disable a certain percentage of neurons at each time step during the training. The idea is that it prevents excessive co-adaptation of neurons that is symptomatic of overfitting.
As a purely hypothetical example, imagine that you had a neuron (or group of neurons) that received signals from a bunch of other neurons in previous layers and fires if it is expected that a creature card will have a mana-producing ability, based on the text preceding the body. The majority of cards we train that match that description are green, and many of those are elves, and many are druids. During training, when we ask it to predict the body text of Llanowar Elves, Joraga Treespeaker, Leaf Gilder, etc. it receives the signals
[creature, green, elf, druid]
and the outcome is "makes mana". At first, it makes mistakes, but overtime, through backpropagation, it connects the dots and it weights the incoming connections such that it will fire when it sees an "elf druid creature that is green". But not all elf druids make mana, and plenty of creatures make mana that aren't elf druids. If it learns an overly specific rules that give it the best performance on the training set, it's less likely to do well on the test set and, in general, less likely to give us interesting cards. What dropout does is randomly disable some neurons and their connections, so instead of seeing all those inputs, instead it sees inputs like
and in doing so, we're forcing the network to separate out the factors that cause creatures to make mana. Green cards can make mana, creatures can make mana, elves can make mana, druids can make mana, etc. In effect, a network with dropout of 25% acts like an optimal combination of "thinner" networks that are 75% of its size.
Obviously there are limits. If the dropout is greater than 50%, then the network can't count on any neuron being reliable, and this can hurt performance.
Dropout is set using the "-dropout" parameter when you call the training script. If you call the training script with "-dropout 0.25", you get 25% dropout, and so forth.
AlphaGo's loss in round 4 is really interesting, as it illuminates the differences in what it means to play and understand a game for human and for a computer, even one backed by a deep NN. (I'm relying heavily on what the pro commentators were saying, especially on the American Go Association channel, so credit where it's due and errors are all mine.)
Thank you for summarizing the commentary! I was too busy to follow the games very closely; what you've shared is very informative for me.
To bring this back to Magic, what it means is that the state of the art represented in AlphaGo would be able to tune the parameters that calculate the worth of a particular game state and even discover deep calculations that human players don't immediately see, but a human player would have the opportunity to probe the AI to find out what it prefers and adjust to match. In a game with such a deep appreciation for the metagame, I don't think an AI is going to be capable of sustained competitive play until there's been further revolutions in AI techniques of comparable magnitude to what brought us from Deep Blue to AlphaGo.
I think that's why Demis Hassabis of DeepMind said that their next target could be something like Starcraft. Mind you, beating the Koreans at all of their favorite games isn't the mission of DeepMind, but a game like Starcraft would be an interesting testbed for metagame analysis. And even if DeepMind doesn't pursue that right away, I assure that others are already thinking along similar lines, just as Facebook and Google's DeepMind have been concurrently working on mastering Go.
I think there are two parts to that analysis. The first is being able to identify a metagame, and the other part is being able to independently come up with possible metagames. These two reinforce each other, of course - generative models and predictive models being two sides of the same coin. I'm convinced that if we can overcome the challenge of metagame analysis for a game with fixed elements like Starcraft, we can do it for a game with evolving elements like Magic.
---
I'm in the process of figuring out how to organize all the image-generating scripts for release. I'd like to set everything up so you can just install the necessary packages, and play with it right away. Some things could be streamlined. I'll see about having all that ready in a day or two.
Could there be a way set up a neural network to play a game of magic? And/or build and it's own deck?
Oh absolutely. Now, to get a bot that can play well, that's another story, and it depends heavily on how we approach the problem and the data that we have. Same goes for deck-building. In many ways, these are highly related topics, but I'll focus on just the gameplay aspect for now.
As we've seen with bots like AlphaGo and Giraffe, it's possible to integrate a neural-network-style approach into a conventional framework. You can have a "game tree" of sorts, where player A makes a move, player B responds to A, A responds to B responding to A, and so on, branching out into all the different possible realities (taking into account randomness and unknown information). A bot looks at each state and says "what is the value of getting to that game state?", and then tries to steer the game towards the path that is most likely to lead to victory. For this, you need an evaluation function, a way of comparing game states, and oftentimes, these are hard-coded.
For example, if you go into the source code for Forge, you'll find all kinds of evaluation rules that the AI. For example,
public static int evaluateCreature(final Card c) {
int value = 100;
if (c.isToken()) {
value = 80; // tokens should be worth less than actual cards
/* ... */
value += power * 15;
value += toughness * 10;
value += c.getCMC() * 5;
// Evasion keywords
if (c.hasKeyword("Flying")) {
value += power * 10;
}
/* ... */
if (c.hasKeyword("Unblockable")) {
value += power * 10;
}
/* ... */
return value;
}
So let's say that the AI has a Doom Blade in hand that it wants to cast, and it sees the opponent has a Storm Crow and a token that is a copy of Tidal Kraken. The bot would like to take out the biggest, most pressing threat.
According to Forge's creature evaluation function, Storm Crow is worth 155 points, and the Tidal Kraken token is worth 270 points. Killing the token would maximize the opponent's losses and minimize the bot's, so the bot kills the Kraken.
In this case, the choice was very clear cut (though some experts might argue that the Storm Crow was inherently more threatening than the Tidal Kraken), but there are limitations to this kind of approach. First, it's not a very empirical approach; we just arbitrarily decided that each point of power was worth 15 points and each point of toughness was worth 10. Second, hand-crafted solutions tend to be very brittle and render the AI unable to respond well to unforeseen situations.
For example, if the AI has an unanswerable Moat on the board, that reduces the Kraken's value to 176 points (I counted), which is still greater than the Crow's. Now, it's possible that there are guards in the code that I didn't see that would prevent the bad decision from being made, but that adds more layers of complexity, and more opportunities for failure.
One way of incorporating machine learning into this process is to replace parts of the evaluation function with learned models (e.g. neural networks). What's nice about these sorts of systems is that their responses have a stronger empirical basis and, in the case of neural networks, can be very flexible in the face of totally unforeseen situations. Of course, adding in magical black boxes can create other issues, like a lack of interpretability. A common example I bring up is when you have a visual question answering system like [url=http://cloudcv.org/vqa/]this one[/url], operating on [url=
http://cloudcv.org/app/media/pictures/vqaDemo/COCO_test2014_0000004518761453763791_36.jpg]this image[/url]. We ask the system the following questions:
Q: "What game is being played?" A: Tennis (confidence 99%)
Q: "What color is the man's shirt?" A: White (confidence 25%, blue in second place with confidence 18%)
Q: "What is the sex of the person?" A: Female (confidence 65%)
Q: "Is his wife cheating on him?" A: No (confidence 96%)
In the first two cases, we get back answers that are correct, and that we can verify using the image. We can even understand why the bot thought the shirt might be blue. In the third case, the bot makes a mistake, but we can forgive it because humans are one of the least sexually dimorphic species on the planet.
But why is the bot so sure that the tennis player's wife is faithful to him? He's always away at tournaments, all he thinks about is improving his game and not his relationship, and she feels oppressed always standing in the shadow of his success. We don't know why the bot is so convinced that she'd be happy in such a terrible relationship. On the bright side, at least it was able to respond with an answer to a question that it was never asked before. That shows flexibility... even if sometimes we don't get interpretability.
---
I need to go back and finish fixing that memory issue with the image generator training process, but it might have to wait. I have a 10 AM meeting tomorrow with someone who has come to give a lecture, a 1:30 PM meeting with an economist interested in doing population projections for our state, and a 3 PM meeting with my team. I have some preparations to make. Oh, and I just got word that another one of my papers was accepted for publication in an international journal, so I have to go celebrate.
But I promise I will return to it! And I'll release all the code needed to do the image generation on your own (including the most recent trained model).
P.S. I hear that AlphaGo lost a round against its human opponent. Interesting!
No, not a fork. What I sampled from was just a private version of mine where I've tinkered with incorporating neural stacks into the architecture. The main repo can be found here (and encoding repo here), and the sampling script "sample_hs_v3.lua" supports field-specific priming like what I've shown you.
Here's a zip file with the encoded corpus in three formats, with cards up to and including Shadows of Innistrad.
And bachelor party? Congratulations!
I guarantee you that that is evidence of the opposite. If you have a phrase in the corpus that is almost always used in a particular context or in a particular way, the network can overfit and assume that it must always be that way, and will memorize the words verbatim. For example, if I try...
I get...
|serpenty serpent||creature||serpent|O|&^^^^/&^^^^|{^^UU^^}|@ can't attack unless defending player controls an island.\{^^UU}: tap target creature and gain control of it until end of turn. that creature gains haste until end of turn.| |serpenty serpent||creature||serpent|O|&^^^^/&^^^^|{^^UU^^}|@ can't attack unless defending player controls an island.\{^UU}: @ can't be blocked this turn.| |serpenty serpent||creature||serpent|O|&^^^^/&^^^^|{^UU^^^}|@ can't attack unless defending player controls an island.\{UU}, sacrifice an island: put a &^/&^ green saproling creature token onto the battlefield.| |serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^^UU^^}|@ can't attack unless defending player controls an island.\when you control no islands, sacrifice @.| |serpenty serpent||creature||serpent|O|&^^^/&^^^|{^UU^^^}|flying\{^^^}, return a land you control to its owner's hand: uncast target spell unless its controller pays {^} for each card in your hand.| |serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^UU^^^}|@ can't attack unless defending player controls an island.\{^UU}: @ can't be blocked this turn.| |serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^UU^^^}|@ can't attack unless defending player controls an island.\{^^UU}: tap target creature an opponent controls.| |serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^UU^^^}|@ can't attack unless defending player controls an island.\{^^^^^^}, T: search your library for a basic land card, reveal it, and put it into your hand. then shuffle your library.| |serpenty serpent||creature||serpent|O|&^^^^/&^^^^|{^^UU^^^}|morph {^^^^^^^}| |serpenty serpent||creature||serpent|O|&^^^^^/&^^^^^|{^^UU^^^}|@ can't attack unless defending player controls an island.\when you control no islands, sacrifice @.\{WW}: @ gains flying until end of turn.| |serpenty serpent||creature||serpent|O|&^^^/&^^^|{^^UU^^}|flying\{^^^}, return a land you control to its owner's hand: uncast target spell unless its controller pays {^} for each card in your hand.|It happens for the same reason that dragon cards usually have flying. It's not that the network knows that dragons are winged creatures and it would make sense for them to fly, it's that the characters "f l y i n g" in the body text often follow the characters "d r a g o n" in the subtype line.
I assure you that, as a rule, I don't touch the cards. Infecting the results with my own ideas would defeat the purpose of the whole experiment, lol. For the same reason, I'm still working out the graphics thing, so I can get art for cards without having to assign art myself. Ran into a small bug earlier in the week with the model, something's not getting loaded properly when I try and restore the model for sampling. I think it's an issue of my own making though. I'll get it resolved.
I should also say that I'm really excited about how far the Hunter project has come along.
EDIT: If you wanted a Turing test for authenticity of cards, suspicious for me would be a card that is significantly longer than about 160 characters or so, because the network hates run-ons. For the same reason, it can't generate complete planeswalkers or sensible flip cards. That and it's unlikely (though not impossible) for the network to produce cards with multiple interconnected or highly related abilities (aside from "add a charge counter"/"remove a charge counter", etc.). For example, Brimaz, King of Oreskos is unlikely to be produced (unless it's overfitting), simply because the network's attention span and its limited comprehension of the semantics of cards pose significant challenges when it comes to sustained, meaningful creation.
Speaking of art, I'm rerunning the training process to condition the generator on the card vectors. Well, part of the card vectors. I left off the body text because that just muddies the waters too much (there's a correlation between "flying" in the bodytext and the artwork, but not so much "enters the battlefield tapped", etc.). I have to make sure I did everything right, do some tests, etc. But I'll let y'all know how it turns out.
Yeah, I can look into that later and let y'all know. And yeah, the distributions that the machine learns and the actual distributions can vary. I mean, within cards, it does a good job, like Dragons and Angels should almost always have flying, but in terms of the distribution of card types, colors, etc. there can be some deviations.
For example, there are more green creature cards than creature cards of any other color, but, when you analyze RoboRosewater's output, you tend to find that all colors are equally likely. RoboRosewater also has a stronger bias towards making 3 CMC creatures than you see in the set of real Magic cards.
Like you said, I think that some creature types get subsumed into others. Like merfolk and vedalken are functionally similar, but there are more merfolk than vedalken, so if it wants to create a Caller of Gales-esque utility creature, it might be biased towards merfolk even when a vedalken would have worked just as well. Upping the temperature can fix that, but it also causes it to make more mistakes as a result of its frantic excitement.
-------
As an aside, I just got done with fielding questions over Skype reagrding population projections using neural networks for the benefit of folks at a US Census Bureau conference. To paraphrase myself: "I know that our techniques must seem like witchcraft to you... and you would be right! It *is* witchcraft. And it's incredibly effective."
On a related note, I retooled my scripts so that I could feed in animations like GIF files.
Forest waterfall -> reconstruction
Scarecrow model from World of Warcraft -> reconstruction
A peaceful scene -> reconstruction
A peaceful scene + vector representation of an image of fire -> the clouds are now a fiery tempest eradicating everything
Simulated rotation of the Girl with the Pearl Earring -> Simulated rotation of the Lampshade Lady.
Still no face on the Lampshade Lady, but as I pointed out before, I messed up the bounding boxes when we trained the last network, and faces got cut out so often that the network seems to strategically place objects in the way of faces in order to avoid having to draw them. That'll get fixed in the next iteration, and we'll also have increased resolution.
Again, the idea here isn't to try and perfectly reconstruct the input, but to use the input as a means of situating ourselves inside the vector space in a region that structurally/semantically resembles the input. The waterfall, for example, isn't the original waterfall, but happens to be another, similar waterfall inside our miniature MTG-art-inspired universe. You can think of these animations as if you were looking through a tiny pinhole into this bizarre little world.
It's fun to play with, because we never trained the system on animations, and it's nice to see that the concept of motion carries over well. That and it's fun to set people, places, and things on fire.
More to come in the near future. Sorry for the wait.
Oh, and I saw a fun paper by Li Fei-Fei's people at Stanford University on real-time neural style transfer and super-resolution. Like, a speed-up by a factor of 1000. Like I said several months ago, these improvements would come along :D. At this point, I'm just waiting for someone to come out with a Google Cardboard app so I can strap it on and see everything as a Van Gogh painting.
@nyrt: Fascinating find! Thank you for sharing!
@Elseleth: I'm excited to see your robo-cube come to fruition.
I could have / should have done that. As written, the code has a 32x32 network configuration and a 64x64 network configuration. I changed it so that I could create an NxN configuration for some value of N that I passed. Easy enough. But yes, recursive deconvolution is a thing, and I should look into it. My first goal right now is to condition the network using tags for cards, so that it learns to relate goblin/sorcery/green card artworks with each other. That way, I can control the output by asking it to make something extra "goblin-y", for example. That increases the odds that the art will make sense in the context that I want to use it. I just about have that ready to go; I'm just waiting on the opportunity to run the code again, and that should be soon. Then I can see about restructuring things to get higher resolution images out.
Not me, that's for sure. I recognize some of the names though, like Grefenstette. It looks like the Google Deepmind folks found a way to improve upon our work. I had recommended the use of a bidirectional LSTM, but I've been too busy to get around to testing that. The use of code compression is novel, however. I'm happy that I was able to attract people with adequate funding/time to investigate this topic in more detail.
I'm gonna have to take a look at their implementation whenever that becomes available. I might e-mail them about that later. The numbers look really good.
Thank you for sharing this! This is most helpful.
EDIT(1): I sent an e-mail to the lead authors with my congratulations and inquiries.
EDIT(2): I had a lovely conversation with the folks at DeepMind. Their work has a lot of fun tricks that I could co-opt for our purposes. They're not sure about a timeline for releasing their source code though, as there's a bureaucratic process for all that. I'd have to re-implement it. Which is fine.
EDIT(3): By the way, when I get the art renders working at higher resolutions (and one way or another, I will eventually), I'm going to have so much fun with it. I've found that, when I choose to guide the art generation process with an example, if it doesn't recognize an object in the scene, it tends to replace it with something else that conforms to the geometry. But it's not an exact match, it has artistic freedom (e.g. adding in feet, tail feathers, a mouth, and a dot for an eye). I think this process closely resembles what surrealist André Breton called "pure psychic automatism".
Oh, and I also attached a sample output that I made for one Thraximundar_ of reddit. The art was created by interpolation/inspiration from the art of semantically similar cards. Bit hard to make out, but like I said, that's one of the things I'm working on. The flavor text I kinda cheated on simply because I produced ten of them and hand-picked the best one; I'll need to work out an automated system for that so I don't contaminate RoboRosewater's cards with my ideas, lol.
Took me longer to reply than I had intended, but hello and welcome! I had a look at your source code. I'm most impressed! I admire very deep and nested procedural generations; I can tell there's a lot of attention paid to detail.
I concur with everything that our resident expert (he's modest) maplesmall has told you. There are challenges with using character-level generative models, and it's not always the best fit for every situation. At the very least, I'm not sure that I'd recommend it as an end-to-end solution for you. However, there are plenty of ways that you could incorporate ML tools and techniques into your generation process (piece-wise).
For example, a lot of the stats can be drawn from learned distributions if needed. That being said, you've spent a lot of time and energy calibrating your stat-purchasing model, and I see no reason to throw that out if it works well for you.
Abilities could be generated in a way similar to what we do, though I might recommend a word-level model rather than our character-level one. Or perhaps even some kind of clause-level model, so it'd be like generating an abstract syntax tree of sorts... but I'm not know whether there are any good implementations of that sort available online.
Now, description text, there's something that you're going to be hard-pressed to do in a purely procedural way (unless you want it to come off as very artificial sounding). A generative architecture like ours could churn out monster description text just fine (and you have plenty of data for that), but it wouldn't deliver what you wanted because what it generated would not be conditioned upon or keyed to the monsters, so you'd get fascinating but arbitrary text attached to your monsters. Instead, you'd want to use a conditional neural language models, like the one used here. But instead of picture in -> text out, it'd be monster in -> text out. Same sort of idea. Of course, the technology for that is still maturing, so the results will probably be something of a mixed bag.
If you'd like, you can PM me or send me an email at rmmilewi (at) gmail (dot) com and we can talk more about this in detail. I can see about directing you towards the resources that you'd need.
By the way, while we're on the subject, Alec Radford just put out a lecture entitled "Deep Advances in Generative Modeling". It's a 40 minute presentation that covers virtually all the algorithms that we've been talking about in this thread and then some. If you skip ahead to 34:50, you can see some unpublished results where they condition an image generator on text.
-----
As for me, I've just about got the art-generation-conditioned-on-card-vectors thing coded, but I'm having to wait to run it. Right now I'm training a bunch of LSTM networks to do population forecasting for my state on behalf of some economists and census folks. But once that's done with, I'll be sure to train a new image generation model. Once I get that working, I can see about writing some scripts that'll integrate the card and art generation processes.
EDIT: On a related note, I just got some new hardware in, so that may mean that I'll be able to run Magic experiments in parallel with others. That'll speed things up, lol.
Another thing to test is to vary the way the representation is organized. The networks I train use a fixed order for all the fields like these:
You can control how the input is organized by passing different parameters to the mtgencode program.
By the way, Master of the Master is... a very unusual card.
Master of the Master
3W
Creature - Human Knight (Rare)
First strike, protection from white and from black.
As long as Master of the Master is on the stack, each creature has "1W, T: This creature deals 1 damage to target creature or player."
2/2
Are there any cards that resemble it?
Anyway, on the subject of dropout, note that overfitting (where the network copies the cards it studies too closely) can come in many forms. For example, this is what I call a pseudo-clone:
Mesa Enchantress
GW
Creature - Human Druid (Rare)
Whenever an enchantment is put into a graveyard from the battlefield, you may put a +1/+1 counter on Mesa Enchantress.
1/2
It has Mesa Enchantress's name and Femeref Enchantress's cost, body, and trigger clause. Just something to look out for when you are comparing results taken from networks trained under different conditions.
It can do either. You can either use an existing checkpoint as a starting point, or you can start all over. To start training from an existing checkpoint, use the "-init_from" parameter followed by the path to the checkpoint that you want to use as a starting point.
The % represents a counter type, which is defined elsewhere in the card. In the case of the card you generated, it should be read as...
Gorgon Camph
2R
Creature - Human Berserker (Common)
W, Sacrifice Gorgon Camph: target creature gets -1/-1 until end of turn.
3R: Put a Fide counter on Gorgon Camph.
2/1
How delightfully bizarre.
Do those directories exist? The cv subdirectory, followed by the custom_format-256 directory? If not, you could create them with the mkdir command. That might be why you are having an issue.
---
Earlier in the week, I said I'd release a Magic-art-image-generation model and script for playing with it. Then I got hit by an avalanche of work. On the bright side, it's Spring break next week, so that frees up my schedule somewhat. I'm going to look into retraining the system to run with tags (images are passed to the network along with tags like "red" or "goblin", so that way the tags give us a more gentle way of controlling the output image rather than operating on the latent space directly.
Oh, and, as usual, the techinques that I've been using are already obsolete (probably). I just read a fascinating paper by Wang and Gupta entitled "Generative Image Modeling using Style and Structure Adversarial Networks". The novel aspect of their work is that they split up the generator network into two parts. The first is a structure generator that creates the geometry of the image (surface normals). Then that geometry is passed to the style generator that handles textures, lighting, appearance, etc.
Right now, with the setup we've got, style and structure are intermixed. That's why when I change the lighting of a scene, the geometry warps and shifts. Wang and Gupta are able to control these two aspects independently.
For the training data, you'd have to have surface normals for the artwork, but that's actually quite feasible. There are trained models out there for predicting the geometry of 2D images (example). Those systems are usually trained on photographs, but I suspect that they could work decently on artwork. Murk Dwellers no, Rafiq of the Many yes.
Not that I'll mess with that at the moment, but it's nice to know that these options exist.
EDIT: Oh, almost forgot! A team from Google made yet another interesting breakthrough in a paper entitled "One-Shot Generalization in Deep Generative Models". One-shot learning is where you have one or two images of a thing and you then have to arrive at a understanding of what that thing is (or in this case, be able to generate new and interesting versions of it). It's sort of a "holy grail" in computer vision. They're one step closer to achieving that. I've attached an example where the system is presented with a strange alphabet that it has never been trained on, and has to learn to write letters in that alphabet from just those single examples. They did another similar test with faces. Just a few faces, make new faces, that sort of thing. And it does decently, actually. Not bad. There's a lot more that needs to be done in this area, but the pay-off could be great. An example application would be to show a picture of an apple to a robot that has never seen apples before and say "This is an apple. Go find me more of these things. Bring them back to me. Kthxbai."
Strange. Well, at least you found a workaround. I'm not sure why it wasn't working for you. Well, the permissions problem I understand, but not the "no results matching query problem."
The numbers themselves aren't especially important, it's just that you need to feed in a checkpoint file, preferably the most recent one, whichever that is. The numbers at the end should tell you the epoch.
Dropout is a regularization technique whereby we randomly disable a certain percentage of neurons at each time step during the training. The idea is that it prevents excessive co-adaptation of neurons that is symptomatic of overfitting.
As a purely hypothetical example, imagine that you had a neuron (or group of neurons) that received signals from a bunch of other neurons in previous layers and fires if it is expected that a creature card will have a mana-producing ability, based on the text preceding the body. The majority of cards we train that match that description are green, and many of those are elves, and many are druids. During training, when we ask it to predict the body text of Llanowar Elves, Joraga Treespeaker,
Leaf Gilder, etc. it receives the signals
[creature, green, elf, druid]
and the outcome is "makes mana". At first, it makes mistakes, but overtime, through backpropagation, it connects the dots and it weights the incoming connections such that it will fire when it sees an "elf druid creature that is green". But not all elf druids make mana, and plenty of creatures make mana that aren't elf druids. If it learns an overly specific rules that give it the best performance on the training set, it's less likely to do well on the test set and, in general, less likely to give us interesting cards. What dropout does is randomly disable some neurons and their connections, so instead of seeing all those inputs, instead it sees inputs like
[creature, green, elf]
[green,elf,druid]
[creature,elf,druid]
[elf,creature]
[creature,green]
[druid]
and in doing so, we're forcing the network to separate out the factors that cause creatures to make mana. Green cards can make mana, creatures can make mana, elves can make mana, druids can make mana, etc. In effect, a network with dropout of 25% acts like an optimal combination of "thinner" networks that are 75% of its size.
Obviously there are limits. If the dropout is greater than 50%, then the network can't count on any neuron being reliable, and this can hurt performance.
Dropout is set using the "-dropout" parameter when you call the training script. If you call the training script with "-dropout 0.25", you get 25% dropout, and so forth.
Thank you for summarizing the commentary! I was too busy to follow the games very closely; what you've shared is very informative for me.
I think that's why Demis Hassabis of DeepMind said that their next target could be something like Starcraft. Mind you, beating the Koreans at all of their favorite games isn't the mission of DeepMind, but a game like Starcraft would be an interesting testbed for metagame analysis. And even if DeepMind doesn't pursue that right away, I assure that others are already thinking along similar lines, just as Facebook and Google's DeepMind have been concurrently working on mastering Go.
I think there are two parts to that analysis. The first is being able to identify a metagame, and the other part is being able to independently come up with possible metagames. These two reinforce each other, of course - generative models and predictive models being two sides of the same coin. I'm convinced that if we can overcome the challenge of metagame analysis for a game with fixed elements like Starcraft, we can do it for a game with evolving elements like Magic.
---
I'm in the process of figuring out how to organize all the image-generating scripts for release. I'd like to set everything up so you can just install the necessary packages, and play with it right away. Some things could be streamlined. I'll see about having all that ready in a day or two.
Oh absolutely. Now, to get a bot that can play well, that's another story, and it depends heavily on how we approach the problem and the data that we have. Same goes for deck-building. In many ways, these are highly related topics, but I'll focus on just the gameplay aspect for now.
As we've seen with bots like AlphaGo and Giraffe, it's possible to integrate a neural-network-style approach into a conventional framework. You can have a "game tree" of sorts, where player A makes a move, player B responds to A, A responds to B responding to A, and so on, branching out into all the different possible realities (taking into account randomness and unknown information). A bot looks at each state and says "what is the value of getting to that game state?", and then tries to steer the game towards the path that is most likely to lead to victory. For this, you need an evaluation function, a way of comparing game states, and oftentimes, these are hard-coded.
For example, if you go into the source code for Forge, you'll find all kinds of evaluation rules that the AI. For example,
So let's say that the AI has a Doom Blade in hand that it wants to cast, and it sees the opponent has a Storm Crow and a token that is a copy of Tidal Kraken. The bot would like to take out the biggest, most pressing threat.
According to Forge's creature evaluation function, Storm Crow is worth 155 points, and the Tidal Kraken token is worth 270 points. Killing the token would maximize the opponent's losses and minimize the bot's, so the bot kills the Kraken.
In this case, the choice was very clear cut (though some experts might argue that the Storm Crow was inherently more threatening than the Tidal Kraken), but there are limitations to this kind of approach. First, it's not a very empirical approach; we just arbitrarily decided that each point of power was worth 15 points and each point of toughness was worth 10. Second, hand-crafted solutions tend to be very brittle and render the AI unable to respond well to unforeseen situations.
For example, if the AI has an unanswerable Moat on the board, that reduces the Kraken's value to 176 points (I counted), which is still greater than the Crow's. Now, it's possible that there are guards in the code that I didn't see that would prevent the bad decision from being made, but that adds more layers of complexity, and more opportunities for failure.
One way of incorporating machine learning into this process is to replace parts of the evaluation function with learned models (e.g. neural networks). What's nice about these sorts of systems is that their responses have a stronger empirical basis and, in the case of neural networks, can be very flexible in the face of totally unforeseen situations. Of course, adding in magical black boxes can create other issues, like a lack of interpretability. A common example I bring up is when you have a visual question answering system like [url=http://cloudcv.org/vqa/]this one[/url], operating on [url=
http://cloudcv.org/app/media/pictures/vqaDemo/COCO_test2014_0000004518761453763791_36.jpg]this image[/url]. We ask the system the following questions:
Q: "What game is being played?" A: Tennis (confidence 99%)
Q: "What color is the man's shirt?" A: White (confidence 25%, blue in second place with confidence 18%)
Q: "What is the sex of the person?" A: Female (confidence 65%)
Q: "Is his wife cheating on him?" A: No (confidence 96%)
In the first two cases, we get back answers that are correct, and that we can verify using the image. We can even understand why the bot thought the shirt might be blue. In the third case, the bot makes a mistake, but we can forgive it because humans are one of the least sexually dimorphic species on the planet.
But why is the bot so sure that the tennis player's wife is faithful to him? He's always away at tournaments, all he thinks about is improving his game and not his relationship, and she feels oppressed always standing in the shadow of his success. We don't know why the bot is so convinced that she'd be happy in such a terrible relationship. On the bright side, at least it was able to respond with an answer to a question that it was never asked before. That shows flexibility... even if sometimes we don't get interpretability.
---
I need to go back and finish fixing that memory issue with the image generator training process, but it might have to wait. I have a 10 AM meeting tomorrow with someone who has come to give a lecture, a 1:30 PM meeting with an economist interested in doing population projections for our state, and a 3 PM meeting with my team. I have some preparations to make. Oh, and I just got word that another one of my papers was accepted for publication in an international journal, so I have to go celebrate.
But I promise I will return to it! And I'll release all the code needed to do the image generation on your own (including the most recent trained model).
P.S. I hear that AlphaGo lost a round against its human opponent. Interesting!