Day 49 is kicking my teeth in. Any suggestions beyond 'take the hills at all costs'
if you have red riding hood, use her as much as you possibly can; same with der freischutz, although hes way harder to use properly.
spread your employees out before arbiter spawns, maybe like 3 in each room at most, as close to the abnormality theyre supposed to work with as possible, because the spikes she summons can stack on top of each other. when arbiter spawns, have about eight people ready to fight her, all in ALEPH ego, and prioritise working on the abnormalities you never want to escape under any circumstances (nothing there, mountain of smiling bodies). you can let breaching abnormalities rampage through the facility on day 49 as long as they dont kill anyone important. however, do NOT fight arbiter unless shes stunned from you undoing her meltdowns of gold, otherwise her pillar attacks will cause more meltdowns. just have your eight guys waiting in a room nearby for her to be stunned, rush in, deal as much damage as possible, then fuck off when she gets up.
i know clockface kills/panics someone when you pause. you are going to need to pause anyway. try to avoid pausing too much, because the amount of employees he kills/panics increases every time you pause, but if you absolutely need to pause (arbiter spawns meltdowns that are too far away, too many dangerous abnormalities breach at once, etc) do it and pray he makes someone panic and not die.
dont worry about resources, or lives lost. day 50 is a victory lap. you can throw away every single employee but as long as you still have one at the end of the day to prevent an autoloss, youre golden. throw absolutely everything at the arbiter and the white ordeals.
if arbiter spawns at the same time as white noon, the runs lost. just restart
3 notes
·
View notes
If you did not already know
AutoLoss
Many machine learning problems involve iteratively and alternately optimizing different task objectives with respect to different sets of parameters. Appropriately scheduling the optimization of a task objective or a set of parameters is usually crucial to the quality of convergence. In this paper, we present AutoLoss, a meta-learning framework that automatically learns and determines the optimization schedule. AutoLoss provides a generic way to represent and learn the discrete optimization schedule from metadata, allows for a dynamic and data-driven schedule in ML problems that involve alternating updates of different parameters or from different loss objectives. We apply AutoLoss on four ML tasks: d-ary quadratic regression, classification using a multi-layer perceptron (MLP), image generation using GANs, and multi-task neural machine translation (NMT). We show that the AutoLoss controller is able to capture the distribution of better optimization schedules that result in higher quality of convergence on all four tasks. The trained AutoLoss controller is generalizable — it can guide and improve the learning of a new task model with different specifications, or on different datasets. …
FPV-TPV
We explore the problem of intersection classification using monocular on-board passive vision, with the goal of classifying traffic scenes with respect to road topology. We divide the existing approaches into two broad categories according to the type of input data: (a) first person vision (FPV) approaches, which use an egocentric view sequence as the intersection is passed; and (b) third person vision (TPV) approaches, which use a single view immediately before entering the intersection. The FPV and TPV approaches each have advantages and disadvantages. Therefore, we aim to combine them into a unified deep learning framework. Experimental results show that the proposed FPV-TPV scheme outperforms previous methods and only requires minimal FPV/TPV measurements. …
Extreme Tensoring
State-of-the-art models are now trained with billions of parameters, reaching hardware limits in terms of memory consumption. This has created a recent demand for memory-efficient optimizers. To this end, we investigate the limits and performance tradeoffs of memory-efficient adaptively preconditioned gradient methods. We propose extreme tensoring for high-dimensional stochastic optimization, showing that an optimizer needs very little memory to benefit from adaptive preconditioning. Our technique applies to arbitrary models (not necessarily with tensor-shaped parameters), and is accompanied by regret and convergence guarantees, which shed light on the tradeoffs between preconditioner quality and expressivity. On a large-scale NLP model, we reduce the optimizer memory overhead by three orders of magnitude, without degrading performance. …
Matryoshka Network
In this paper, we develop novel, efficient 2D encodings for 3D geometry, which enable reconstructing full 3D shapes from a single image at high resolution. The key idea is to pose 3D shape reconstruction as a 2D prediction problem. To that end, we first develop a simple baseline network that predicts entire voxel tubes at each pixel of a reference view. By leveraging well-proven architectures for 2D pixel-prediction tasks, we attain state-of-the-art results, clearly outperforming purely voxel-based approaches. We scale this baseline to higher resolutions by proposing a memory-efficient shape encoding, which recursively decomposes a 3D shape into nested shape layers, similar to the pieces of a Matryoshka doll. This allows reconstructing highly detailed shapes with complex topology, as demonstrated in extensive experiments; we clearly outperform previous octree-based approaches despite having a much simpler architecture using standard network components. Our Matryoshka networks further enable reconstructing shapes from IDs or shape similarity, as well as shape sampling. … https://analytixon.com/2022/11/16/if-you-did-not-already-know-1887/?utm_source=dlvr.it&utm_medium=tumblr
0 notes
If you think Verlisify is bad at Pokemon, his Yu-Gi-Oh strategies were scummy af. Supposedly, he was a level 2 judge and knew a lot of shady rulings so he always fell back on "Dropoff OTK". Dropoff/Drastic Dropoff made your opponent immediately discard the card they drew during the draw phase (or drawing from a card effect for Drastic Dropoff) , so if you shuffle your hand before either Dropoff is activated without discarding, its an irreparable game state resulting in autoloss. Its bullshit
Holy shit what a fucking asshole. But honestly, I’m not surprised. The kinds of people who resort to that strategy are the kinds of people who NEED it because they’re too garbage at the game to win otherwise, and if you know anything about Verlis and how he is at comp, he is a fucking dumpster fire of a bad player. Watching his videos will make you WORSE at Pokemon
If you want people who you can watch to get better, PokeaimMD for Smogon and WolfeGlickeVGC for VGC, with the latter being a world champion.
1 note
·
View note
My Run at Richmond Regionals/Day 2 Cup Report
I played Zoroark Garbodor in Richmond last weekend, and things were...weird. So all in all, just another day of Expanded. I went 3-4-2 day one, a worse day than Greensboro last season, however, I do believe that I improved in many ways despite a worse finish (more on that later). Let’s have a look at what went wrong!
R1 vs. Archie’s Night March LWT
This was such a cool deck! They used Archie’s Ace in the Hole with Swampert to give them more draw to improve control matchups. In game 1, I played a Karen to disrupt his damage output and he responded with double compressor, power draw and sycamore to build right back up on a one-shot and took the game fairly easily. Game 2 was a different story, Karen actually stuck and I just took my 6 prizes without any difficulty, which is really the only way to win this matchup. We got to game 3, he went first, played power plant, let loose’d me into an unplayable hand, I got basics down and then time was called. Super lucky to have tied that one, given the hand I got in that third game.
R2 vs. Archiestoise LL
This is the most polarized matchup ZoroGarb has. If they stumble at all turn 1, we can just win. If they get a cross division off, we just lose. Neither of those things happened here though! I whiffed energy the entire round and by the time I was hitting energy, it didn’t matter as the board state was too far gone.
R3 vs. MewGardy LWW
MewGardy is just a really bad deck. I prized double trubbish game one and couldn’t get there with Zoroarks alone, but as soon as I had access to all my pokemon in the second two games it was pretty easy. This would’ve been a tie had my opponent not scooped on turn 2 of time, seeing that his board was completely lost. S/o to him for being friendly!
R4 vs. ZoroToad LL
This. Matchup. Sucks. I was item locked the entierty of both games, with sudo down and no way to take prizes. Probably just an autoloss. In game 1, I saw the toad and assumed it was a control variant, which we can beat. But when they’re prize racing with LaserBank and Quaking Punch, our options are limited. ZoroToad doesn’t have many favorable matchups in this format, I would even go so far as to say that it’s just bad. But it definitely beats ZoroGarb most of the time.
R5 vs. MewGardy WW
My opponent had a rough start both games and by the time he was attacking, trashalanche was one-shotting everything. Round was over in a little under 20 minutes.
R6 vs. Archiestoise WW
The matchup went the other way this time. My opponent had decent turn ones both games but my deck actually functioned properly and Cross Division never happened. Great games with a super friendly opponent, though.
R7 vs. Trev LL
This used to be a free matchup last season. Now, however, they play 4 Power Plant. It’s more like 50/50 now, but if they brick you off of Let Loose + Plant there’s just nothing you can do. I got a board set up but never got to a point where I could attack and whiffed evolutions all both games. Nothing I think I could’ve done differently, but it all felt so bad. This was the round that took us out of contention for day 2.
R8 vs. Durant WLT
With Oranguru, this matchup is pretty favored. Took game one through Resource Management and KOing Ants. Game 2 I wasn’t able to develop to a point where I could be aggressive enough to take out ants often. Eventually we decked out after some Handiwork double heads. Time was called in game three, I had 2 prizes left by the end. Neither of us wanted to tie but both of us were trying to hit points, we talked with a judge about how else we could resolve it. Going by prizes isn’t really fair when it’s a mill strategy so we just agreed to draw. Thinking about it, I wonder if there is a mathematical way to equate cards left in deck to prize cards and use that to decide. Either way, this took us out of contention for points.
R9 vs. Archiestoise WLL
This is the only round I really felt like I was playing sub-optimally. I was tired, I couldn’t get anything no matter what happened this round, I just wanted to leave. I was basically brain dead and on autopilot the whole way through. I don’t even remember how I got the first game, it just sorta happened. He got Cross Division off Games 2 and 3 and there was just nothing happening for me. Oh well.
Looking back through all of the rounds I would do a few things differently.
I would’ve cut the girafarig. It’s good, but I never hit the matchups it’s good in. Cobalion GX would’ve been way more helpful to stop cross division.
In a more broad sense though, I think I should’ve just played Zoroark Control. I’m comfortable enough with the deck, the only reason I didn’t play it was that I didn’t think I could play it perfectly for 9 rounds. In retrospect, I think I still would’ve had better chances with it than what I played.
On the positive side, I played much better than I did in Greensboro. I remember making small mistakes and throwing matches last time. Here, I think my play improved, but everything else got worse. My matchups were horrendous, I drew poorly off of Let Loose + Plant. A lot went wrong.
I can place an amount of blame on my preparation, for sure. I think the list could’ve been slightly better, and that’s definitely my fault. But at the same time, my list was geared to beat the decks I was expecting, and I didn’t come across as many of those as I would’ve liked.
After that pretty awful day 1, I played in the League Cup on day 2 with a different version of the deck. Playing cards like Bodybuilding Dumbbells and a heavier Parallel city count. I no longer have the exact list, but you aren’t missing much. Probably a little worse than what I played in the main event. But we’ll have a look at those rounds too, just for fun.
R1 vs. TurboDark W
I played against a player that attends tournaments in my area, which was rather nice. This was the first and only time I played against Dark the whole weekend and it was one of the most hyped decks going into the tournament. With Parallel and Sudowoodo, the matchup is pretty much just free.Keep control of their board and trashalanche until they lose.
R2 vs. Pikarom L
Half-bricked at the start of this game, had to bench more Lele and Dedenne than I would’ve liked. Game lasted a total of three turns, and two attacks. Full Blitz, Tag Bolt GG.
R3 vs. ZoroGarb T
This round didn’t go too well. He played cards that I wasn’t expecting, namely Acerola and Professor Kukui. We both kept Sudowoodo down the entire game. Trading two shots with Acerola inbetween for 30 minutes, then I used Tapu Cure GX to heal all the damage off my board in turn 2 of time and that was it.
R4 vs. Archiestoise W
My opponent missed Archie’s turn 1, 2 and 3. I don’t really have anything else to say here.
R5 vs. Archiestoise L
We both had horrible starts, but Cross Division made mine way worse. It was a close game in all honesty, but he got rid of all our Trubbish and we just couldn’t do anything from there.
R6 vs. Green’s Greninja BREAK L
Oh dear. This was horrid. No Trade, they limited their items too well for us to one shot anything. Really missed having Sky Field in this one.
R7 vs. Keldeo GX/Hoopa W
They played 0 outs to Garbotoxin, Garb just won the game outright.
I played 16 rounds of Zoroark Garbodor in it’s various forms this weekend! The deck is pretty busted, but the deck just can’t beat a lot of things. Let Loose + Power Plant can make any otherwise good matchup unfavored, we can lose to a lot of very specific random stuff.
Going forward, I think the deck actually gets better! The matchup against the 1st place EggRow deck isn’t too bad with double Klefki and some more hand disruption. For Portland, I would definitely be giving this deck a look! I will leave you all with a warning, however. If you want to play this deck, test it thoroughly. Know each matchup, know how the deck tends to run, always have a plan. The deck has a lot of options and isn’t for someone who wants to just do the same thing every game. For that, I’ll refer you to Turbo Dark.
I’m disappointed in myself for my poor performance and I’m very unhappy with how the tournament went for me. I understand that not everything that happened was in my control, and I’m very happy with how I played for most of the weekend, so I hesitate to call the weekend a complete failure. But I definitely have much to improve on and will be giving my all to the game from now leading up to whatever regionals I’ll be attending next.
Thanks for reading!
0 notes