Meta tags:
author=Micah Goldblum;
description=Micah Goldblum. Math and ML. ;

Headings (most frequently used words):

micah, goldblum, selected, publications,

Text of the page (most frequently used words):
and (107), the (86), for (49), that (42), learning (39), models (35), with (23), are (23), neural (23), micah (20), this (20), #goldblum (19), networks (19), tasks (18), training (18), html (16), abs (16), problems (16), can (16), from (15), systems (14), reasoning (14), data (13), our (13), tom (13), goldstein (13), small (13), batch (13), model (13), visual (12), which (12), image (11), datasets (11), large (11), trained (11), robust (11), work (10), find (10), these (10), conference (10), adversarial (10), performance (10), such (10), machine (10), vision (9), international (9), processing (9), accuracy (9), more (9), meta (9), language (9), high (8), test (8), recognition (8), few (8), generalization (8), shot (8), cot (8), representations (7), scale (7), often (7), show (7), all (7), information (7), classification (7), while (7), diffusion (7), marginal (7), likelihood (7), llms (7), sizes (7), size (7), low (6), deep (6), computer (6), have (6), generated (6), methods (6), face (6), solve (6), search (6), hard (6), advances (6), neurips (6), distillation (6), robustness (6), when (6), adversarially (6), state (6), not (6), perform (6), using (6), architectures (6), set (6), text (6), questions (6), dimension (5), better (5), develop (5), iclr (5), simple (5), inference (5), student (5), introduce (5), benchmarks (5), recent (5), network (5), icml (5), other (5), both (5), andrew (5), gordon (5), wilson (5), bias (5), across (5), range (5), biases (5), pretrained (5), llm (5), over (5), free (5), complexity (5), benchmark (5), livebench (5), widely (4), conventional (4), common (4), images (4), way (4), however (4), release (4), humans (4), difficult (4), harder (4), examples (4), has (4), same (4), time (4), algorithms (4), recurrent (4), complex (4), paper (4), how (4), even (4), ard (4), standard (4), 2020 (4), about (4), wide (4), between (4), feature (4), achieve (4), why (4), well (4), understanding (4), than (4), choice (4), gradient (4), without (4), hyperparameter (4), including (4), hyperparameters (4), based (4), backbones (4), supervised (4), pretraining (4), human (4), binoculars (4), fine (4), tuning (4), optimizer (4), contamination (4), new (4), contains (4), zebra (4), natural (3), dimensional (3), despite (3), success (3), role (3), intrinsic (3), number (3), 2021 (3), its (3), facial (3), increasingly (3), been (3), existing (3), detection (3), easy (3), use (3), valeriia (3), cherepanova (3), strategies (3), learned (3), mazes (3), much (3), larger (3), behavior (3), through (3), their (3), during (3), domains (3), avi (3), schwarzschild (3), furong (3), huang (3), algorithm (3), producing (3), like (3), architecture (3), terms (3), art (3), liam (3), fowl (3), evaluate (3), theory (3), practical (3), jonas (3), geiping (3), furthermore (3), many (3), previous (3), highly (3), learners (3), noise (3), generative (3), community (3), 2023 (3), observations (3), bayesian (3), comparison (3), first (3), provide (3), works (3), mitigation (3), pre (3), make (3), backbone (3), diverse (3), research (3), ssl (3), propose (3), any (3), rate (3), lunch (3), theorems (3), require (3), designed (3), seemingly (3), accumulation (3), fixed (3), math (3), interleaved (3), dataset (3), university (3), 2026 (2), structure (2), estimation (2), tools (2), popular (2), investigate (2), indeed (2), very (2), additionally (2), easier (2), learn (2), generalize (2), process (2), zhu (2), deployed (2), alike (2), typically (2), built (2), social (2), media (2), full (2), entire (2), effective (2), against (2), below (2), john (2), dickerson (2), attacks (2), but (2), may (2), still (2), thinking (2), longer (2), easily (2), extend (2), problem (2), instances (2), computation (2), limited (2), steps (2), simply (2), demonstrate (2), algorithmic (2), chess (2), eitan (2), borgnia (2), knowledge (2), vulnerable (2), only (2), second (2), onto (2), also (2), superior (2), experiments (2), finally (2), aaai (2), assumptions (2), held (2), practitioners (2), kernel (2), setting (2), truth (2), produce (2), extractors (2), known (2), several (2), hypotheses (2), simultaneously (2), renkun (2), requires (2), sets (2), called (2), thoroughly (2), moreover (2), method (2), achieves (2), imagenet (2), approach (2), adding (2), degradation (2), deterministic (2), generalized (2), into (2), question (2), arbitrary (2), hamid (2), kazemi (2), generating (2), prior (2), automatically (2), sensitive (2), revisit (2), 2022 (2), outstanding (2), award (2), sanae (2), lotfi (2), safety (2), yet (2), they (2), exhibit (2), wisdom (2), dictates (2), approaches (2), following (2), fairness (2), suite (2), two (2), most (2), samuel (2), dooley (2), colin (2), white (2), fairer (2), randomly (2), initialized (2), was (2), convolutional (2), battle (2), bob (2), self (2), stable (2), ranging (2), analysis (2), apples (2), similarly (2), sized (2), future (2), modern (2), thought (2), capable (2), spotting (2), sources (2), chatgpt (2), 2024 (2), individual (2), inductive (2), generate (2), kolmogorov (2), notably (2), particular (2), variety (2), automated (2), off (2), down (2), train (2), vanilla (2), sgd (2), building (2), multiple (2), similar (2), 2025 (2), updated (2), answers (2), challenging (2), spanning (2), versions (2), source (2), will (2), jain (2), yann (2), lecun (2), willie (2), neiswanger (2), multimodal (2), quality (2), where (2), scientific (2), wang (2), publications (2), received (2), current (2), columbia (2), copyright, powered, theme, hosted, github, pages, folio, jekyll, believed, exhibits, dimensionality, pixel, idea, underlies, intuition, remarkable, apply, relative, pixels, solving, along, technique, validating, synthetic, gans, allowing, actively, manipulate, controlling, generation, phillip, pope, chen, ahmed, abdelkader, impact, private, corporations, government, agencies, contractors, consumer, services, mass, surveillance, programs, scraping, profiles, user, perturbations, proposed, bypassing, fail, commercial, apis, own, filter, accounts, pipeline, demonstrably, industrial, grade, pipelines, include, databases, webtool, significantly, degrades, amazon, rekognition, microsoft, azure, api, reducing, each, harrison, foley, shiyuan, duan, gavin, taylor, lowkey, leveraging, protect, users, powerful, machines, pattern, possess, ability, extrapolate, example, person, who, techniques, spending, computers, achieved, arbitrarily, cost, contrast, sequential, computing, budget, feed, forward, depth, extending, accommodate, performing, additional, recurrences, prefix, sum, three, able, abilities, arjun, gupta, uzi, vishkin, you, generalizing, studies, transfers, teacher, amount, inherited, distilled, clean, distilling, addition, passes, decisively, outperform, identical, surpassing, adapt, fast, accelerated, proceedings, artificial, intelligence, soheil, feizi, empirically, theorists, prove, widespread, existence, suboptimal, local, minima, loss, landscape, norm, parameters, optimal, resnets, conform, theories, tangent, interaction, skip, connections, normalization, plays, rank, does, correlate, michael, moeller, backpropaganda, empirical, investigation, literature, rich, little, resulting, underlying, mechanics, difference, classically, doing, verify, regularizer, boosts, routines, cases, routine, outperforms, running, order, magnitude, faster, steven, reich, unraveling, computationally, expensive, procedures, hand, goal, querying, causes, vulnerability, far, mini, cifar, transfer, involve, transform, gaussian, restoration, operator, inverts, observe, strongly, dependent, fact, family, constructed, varying, completely, degradations, blur, masking, update, rules, underlie, create, fully, calls, relies, either, langevin, dynamics, variational, paves, invert, processes, arpit, bansal, hong, min, chu, jie, cold, inverting, transforms, compare, entirely, consistent, aka, evidence, represents, probability, provides, distinctive, foundational, encoding, occam, razor, although, observed, overfit, limitations, discrete, investigated, appealing, properties, constraints, hypothesis, testing, then, highlight, conceptual, issues, proxy, namely, negatively, correlated, implications, lead, underfitting, overfitting, partial, remedy, conditional, aligned, practically, valuable, pavel, izmailov, gregory, benton, selection, critical, applications, law, enforcement, socio, demographic, dimensions, gender, race, arise, biased, consequence, largely, focused, penalties, prevent, effecting, post, predictions, debias, them, shown, discover, actually, inherent, themselves, reframing, conduct, jointly, outputs, pareto, dominate, margins, used, identification, celeba, vggface2, attributes, rhea, sukthanker, frank, hutter, rethinking, extractor, years, ago, default, option, past, seen, emergence, countless, various, abundance, led, increases, informed, decisions, choose, makes, benchmarking, those, via, object, ood, sheds, light, promising, directions, advance, illuminating, strengths, weakness, comprehensive, conducted, 1500, runs, transformers, vits, fashion, best, among, consider, comparisons, competitive, indicating, should, advanced, hossein, souri, manli, shu, viraj, uday, prabhu, gowthami, somepalli, prithvijit, chattopadhyay, adrien, bardes, mark, ibrahim, judy, hoffman, rama, chellappa, detecting, behaviors, score, contrasting, closely, related, accurate, separating, mechanism, novel, detector, calculations, pair, specific, modifications, comprehensively, varied, situations, document, types, detects, samples, false, positive, being, abhimanyu, hans, aniruddha, saha, zero, learner, exactly, average, uniform, distribution, accordingly, referenced, support, notion, specially, tailored, virtually, uniformly, sampled, real, world, disproportionately, argue, share, preference, formalized, domain, compress, unrelated, prefer, sequences, whereas, indicate, specialized, explain, intervention, picking, appropriately, labeled, scarce, plentiful, single, justify, trend, unifying, disparate, marc, finzi, keefer, rowan, theorem, unstable, motivating, trades, proportional, increase, decrease, smaller, one, rule, scaling, adam, rather, holding, decay, moment, hold, half, life, tokens, stably, consistently, choices, equal, per, flop, enable, momentum, storing, results, recommendations, selecting, further, recommend, unless, devices, replicas, combined, benefits, maintaining, memory, footprint, lora, martin, marek, aditya, somasundaram, wasteful, wherein, ends, newer, documented, obstacle, fair, evaluation, quickly, render, obsolete, mitigate, crowdsource, prompts, evaluations, judges, significant, break, scoring, immune, pitfalls, judging, crowdsourcing, frequently, scores, according, objective, ground, values, coding, instruction, recently, released, competitions, arxiv, papers, news, articles, big, bench, amps, ifeval, prominent, closed, dozens, open, 110b, top, achieving, code, added, monthly, basis, distinguish, capabilities, improve, welcome, engagement, collaboration, expanding, manley, roberts, arka, pal, ben, feuer, siddhartha, ravid, shwartz, ziv, neel, khalid, saifullah, siddartha, naidu, chinmay, hegde, rely, aids, diagrams, sketches, tackling, teaching, adopt, chain, main, challenges, weak, shelf, hinders, reinforcement, lack, 182, 384, traces, distinct, specifically, natively, emphasize, four, categories, sketching, especially, geometry, physics, jigsaw, puzzles, multi, hop, embodied, robot, planning, logic, strategic, games, anole, yields, improvement, gains, vlm, bagel, produces, chains, underscoring, effectiveness, advancing, ang, charles, deqing, kaiyu, yue, zikui, cai, bill, ollie, liu, peng, guo, selected, portfolio, includes, security, privacy, before, position, postdoctoral, fellow, york, working, mathematics, maryland, worked, wojciech, czaja, assistant, professor, focuses, applied, fundamental, science, mathematical, google, scholar, dot, edu, trying, understand, toggle, navigation,

Text of the page (random words):
ngly unrelated domains our experiments show that pre trained and even randomly initialized language models prefer to generate low complexity sequences whereas no free lunch theorems seemingly indicate that individual problems require specialized learners we explain how tasks that often require human intervention such as picking an appropriately sized model when labeled data is scarce or plentiful can be automated into a single learning algorithm these observations justify the trend in deep learning of unifying seemingly disparate problems with an increasingly small set of machine learning models spotting llms with binoculars zero shot detection of machine generated text abhimanyu hans avi schwarzschild valeriia cherepanova hamid kazemi aniruddha saha micah goldblum jonas geiping and tom goldstein international conference on machine learning icml 2024 abs html detecting text generated by modern large language models is thought to be hard as both llms and humans can exhibit a wide range of complex behaviors however we find that a score based on contrasting two closely related language models is highly accurate at separating human generated and machine generated text based on this mechanism we propose a novel llm detector that only requires simple calculations using a pair of pre trained llms the method called binoculars achieves state of the art accuracy without any training data it is capable of spotting machine text from a range of modern llms without any model specific modifications we comprehensively evaluate binoculars on a number of text sources and in varied situations over a wide range of document types binoculars detects over 90 of generated samples from chatgpt and other llms at a false positive rate of 0 01 despite not being trained on any chatgpt data battle of the backbones a large scale comparison of pretrained models across computer vision tasks micah goldblum hossein souri renkun ni manli shu viraj uday prabhu gowthami somepalli prithvijit chattopadhyay adrien bardes mark ibrahim judy hoffman rama chellappa andrew gordon wilson and tom goldstein advances in neural information processing systems neurips 2023 abs html neural network based computer vision systems are typically built on a backbone a pretrained or randomly initialized feature extractor several years ago the default option was an imagenet trained convolutional neural network however the recent past has seen the emergence of countless backbones pretrained using various algorithms and datasets while this abundance of choice has led to performance increases for a range of systems it is difficult for practitioners to make informed decisions about which backbone to choose battle of the backbones bob makes this choice easier by benchmarking a diverse suite of pretrained models including vision language models those trained via self supervised learning and the stable diffusion backbone across a diverse set of computer vision tasks ranging from classification to object detection to ood generalization and more furthermore bob sheds light on promising directions for the research community to advance computer vision by illuminating strengths and weakness of existing approaches through a comprehensive analysis conducted on more than 1500 training runs while vision transformers vits and self supervised learning ssl are increasingly popular we find that convolutional neural networks pretrained in a supervised fashion on large training sets still perform best on most tasks among the models we consider moreover in apples to apples comparisons on the same architectures and similarly sized pretraining datasets we find that ssl backbones are highly competitive indicating that future works should perform ssl pretraining with advanced architectures and larger pretraining datasets rethinking bias mitigation fairer architectures make for fairer face recognition samuel dooley rhea sukthanker john p dickerson colin white frank hutter and micah goldblum advances in neural information processing systems neurips 2023 abs html face recognition systems are widely deployed in safety critical applications including law enforcement yet they exhibit bias across a range of socio demographic dimensions such as gender and race conventional wisdom dictates that model biases arise from biased training data as a consequence previous works on bias mitigation largely focused on pre processing the training data adding penalties to prevent bias from effecting the model during training or post processing predictions to debias them yet these approaches have shown limited success on hard problems such as face recognition in our work we discover that biases are actually inherent to neural network architectures themselves following this reframing we conduct the first neural architecture search for fairness jointly with a search for hyperparameters our search outputs a suite of models which pareto dominate all other high performance architectures and existing bias mitigation methods in terms of accuracy and fairness often by large margins on the two most widely used datasets for face identification celeba and vggface2 furthermore these models generalize to other datasets and sensitive attributes bayesian model selection the marginal likelihood and generalization sanae lotfi pavel izmailov gregory benton micah goldblum and andrew gordon wilson international conference on machine learning icml outstanding paper award 2022 abs html how do we compare between hypotheses that are entirely consistent with observations the marginal likelihood aka bayesian evidence which represents the probability of generating our observations from a prior provides a distinctive approach to this foundational question automatically encoding occam s razor although it has been observed that the marginal likelihood can overfit and is sensitive to prior assumptions its limitations for hyperparameter learning and discrete model comparison have not been thoroughly investigated we first revisit the appealing properties of the marginal likelihood for learning constraints and hypothesis testing we then highlight the conceptual and practical issues in using the marginal likelihood as a proxy for generalization namely we show how marginal likelihood can be negatively correlated with generalization with implications for neural architecture search and can lead to both underfitting and overfitting in hyperparameter learning we provide a partial remedy through a conditional marginal likelihood which we show is more aligned with generalization and practically valuable for large scale hyperparameter learning such as in deep kernel learning cold diffusion inverting arbitrary image transforms without noise arpit bansal eitan borgnia hong min chu jie s li hamid kazemi furong huang micah goldblum jonas geiping and tom goldstein advances in neural information processing systems neurips 2023 abs html standard diffusion models involve an image transform adding gaussian noise and an image restoration operator that inverts this degradation we observe that the generative behavior of diffusion models is not strongly dependent on the choice of image degradation and in fact an entire family of generative models can be constructed by varying this choice even when using completely deterministic degradations e g blur masking and more the training and test time update rules that underlie diffusion models can be easily generalized to create generative models the success of these fully deterministic models calls into question the community s understanding of diffusion models which relies on noise in either gradient langevin dynamics or variational inference and paves the way for generalized diffusion models that invert arbitrary processes adversarially robust few shot learning a meta learning approach micah goldblum liam fowl and tom goldstein advances in neural information processing systems neurips 2020 abs html previous work on adversarially robust neural networks for image classification requires large training sets and computationally expensive training procedures on the other hand few shot learning methods are highly vulnerable to adversarial examples the goal of our work is to produce networks which both perform well at few shot classification tasks and are simultaneously robust to adversarial examples we develop an algorithm called adversarial querying aq for producing adversarially robust meta learners and we thoroughly investigate the causes for adversarial vulnerability moreover our method achieves far superior robust performance on few shot image classification tasks such as mini imagenet and cifar fs than robust transfer learning unraveling meta learning understanding feature representations for few shot tasks micah goldblum steven reich liam fowl renkun ni valeriia cherepanova and tom goldstein international conference on machine learning icml 2020 abs html meta learning algorithms produce feature extractors which achieve state of the art performance on few shot classification while the literature is rich with meta learning methods little is known about why the resulting feature extractors perform so well we develop a better understanding of the underlying mechanics of meta learning and the difference between models trained using meta learning and models which are trained classically in doing so we introduce and verify several hypotheses for why meta learned models perform better furthermore we develop a regularizer which boosts the performance of standard training routines for few shot classification in many cases our routine outperforms meta learning while simultaneously running an order of magnitude faster truth or backpropaganda an empirical investigation of deep learning theory micah goldblum jonas geiping avi schwarzschild michael moeller and tom goldstein international conference on learning representations iclr 2020 abs html we empirically evaluate common assumptions about neural networks that are widely held by practitioners and theorists alike in this work we 1 prove the widespread existence of suboptimal local minima in the loss landscape of neural networks and we use our theory to find examples 2 show that small norm parameters are not optimal for generalization 3 demonstrate that resnets do not conform to wide network theories such as the neural tangent kernel and that the interaction between skip connections and batch normalization plays a role 4 find that rank does not correlate with generalization or robustness in a practical setting adversarially robust distillation micah goldblum liam fowl soheil feizi and tom goldstein proceedings of the aaai conference on artificial intelligence aaai 2020 abs html knowledge distillation is effective for producing small high performance neural networks for classification but these small networks are vulnerable to adversarial attacks this paper studies how adversarial robustness transfers from teacher to student during knowledge distillation we find that a large amount of robustness may be inherited by the student even when distilled on only clean images second we introduce adversarially robust distillation ard for distilling robustness onto student networks in addition to producing small models with high test accuracy like conventional distillation ard also passes the superior robustness of large networks onto the student in our experiments we find that ard student models decisively outperform adversarially trained networks of identical architecture in terms of robust accuracy surpassing state of the art methods on standard robustness benchmarks finally we adapt recent fast adversarial training methods to ard for accelerated robust distillation can you learn an algorithm generalizing from easy to hard problems with recurrent networks avi schwarzschild eitan borgnia arjun gupta furong huang uzi vishkin micah goldblum and tom goldstein advances in neural information processing systems neurips 2021 abs html deep neural networks are powerful machines for visual pattern recognition but reasoning tasks that are easy for humans may still be difficult for neural models humans possess the ability to extrapolate reasoning strategies learned on simple problems to solve harder examples often by thinking for longer for example a person who has learned to solve small mazes can easily extend the very same search techniques to solve much larger mazes by spending more time in computers this behavior is often achieved through the use of algorithms which scale to arbitrarily hard problem instances at the cost of more computation in contrast the sequential computing budget of feed forward neural networks is limited by their depth and networks trained on simple problems have no way of extending their reasoning to accommodate harder problems in this work we show that recurrent networks trained to solve simple problems with few recurrent steps can indeed solve much more complex problems simply by performing additional recurrences during inference we demonstrate this algorithmic behavior of recurrent networks on prefix sum computation mazes and chess in all three domains networks trained on simple problem instances are able to extend their reasoning abilities at test time simply by thinking for longer lowkey leveraging adversarial attacks to protect social media users from facial recognition valeriia cherepanova micah goldblum harrison foley shiyuan duan john p dickerson gavin taylor and tom goldstein international conference on learning representations iclr 2021 abs html facial recognition systems are increasingly deployed by private corporations government agencies and contractors for consumer services and mass surveillance programs alike these systems are typically built by scraping social media profiles for user images adversarial perturbations have been proposed for bypassing facial recognition systems however existing methods fail on full scale systems and commercial apis we develop our own adversarial filter that accounts for the entire image processing pipeline and is demonstrably effective against industrial grade pipelines that include face detection and large scale databases additionally we release an easy to use webtool that significantly degrades the accuracy of amazon rekognition and the microsoft azure face recognition api reducing the accuracy of each to below 1 the intrinsic dimension of images and its impact on learning phillip pope chen zhu ahmed abdelkader micah goldblum and tom goldstein international conference on learning representations iclr 2021 abs html it is widely believed that natural image data exhibits low dimensional structure despite the high dimensionality of conventional pixel representations this idea underlies a common intuition for the remarkable success of deep learning in computer vision in this work we apply dimension estimation tools to popular datasets and investigate the role of low dimensional structure in deep learning we find that common natural image datase...