Overparameterized ML models, including neural networks, typically induce underdetermined training objectives with multiple global minima. The implicit bias refers to the limiting global minimum that is attained by a common optimization algorithm, such as gradient descent (GD). In this paper, we characterize the implicit bias of GD for training a shallow ReLU model with the squared loss on high-dimensional random features. Prior work (Vardi and Shamir, 2021) showed that the implicit bias does not exist in the worst-case, or corresponds exactly to the minimum-$\ell_2$-norm interpolating solution under exactly orthogonal data (Boursier et al., 2022). Our work interpolates between these two extremes and shows that, for sufficiently high-dimensional random data, the implicit bias approximates the minimum-$\ell_2$-norm solution with high probability with a gap on the order $\Theta(\sqrt{n/||\lambda||_1})$, where $n$ is the number of training examples and $\lambda$ denotes the spectrum of the data covariance matrix. Our results are obtained through a novel primal-dual analysis that carefully tracks the evolution of predictions, data-span coefficients, as well as their interactions, and show that the ReLU activation pattern quickly stabilizes with high probability over random data.

Modern machine learning methods have recently demonstrated remarkable capability to generalize under task shift, where latent knowledge is transferred to a different, often more difficult, task under a similar data distribution. We investigate this phenomenon in an overparameterized linear regression setting where the task shifts from classification during training to regression during evaluation. In the zero-shot case, wherein no regression data is available, we prove that task shift is impossible in both sparse signal and random signal models for any Gaussian covariate distribution. In the few-shot case, wherein limited regression data is available, we propose a simple postprocessing algorithm which asymptotically recovers the ground-truth predictor. Our analysis leverages a fine-grained characterization of individual parameters arising from minimum-norm interpolation which may be of independent interest. Our results show that while minimum-norm interpolators for classification cannot transfer to regression a priori, they experience surprisingly structured attenuation which enables successful task shift with limited additional data.

Suppose we observe a trajectory of length $n$ from an exponentially $\alpha$-mixing stochastic process over a finite but potentially large state space. We consider the problem of estimating the probability mass placed by the stationary distribution of any such process on elements that occur with a certain frequency in the observed sequence. We estimate this vector of probabilities in total variation distance, showing universal consistency in $n$ and recovering known results for i.i.d. sequences as special cases. Our proposed methodology—implementable in linear time—carefully combines the plug-in (or empirical) estimator with a recently-proposed modification of the Good–Turing estimator called WingIt, which was originally developed for Markovian sequences. En route to controlling the error of our estimator, we develop new performance bounds on WingIt and the plug-in estimator for exponentially $\alpha$-mixing stochastic processes. Importantly, the extensively used method of Poissonization can no longer be applied in our non i.i.d. setting, and so we develop complementary tools—including concentration inequalities for a natural self-normalized statistic of mixing sequences—that may prove independently useful in the design and analysis of estimators for related problems. Simulation studies corroborate our theoretical findings.

We provide a unified framework that applies to a general family of convex losses across binary and multiclass settings in the overparameterized regime to approximately characterize the implicit bias of gradient descent in closed form. Specifically, we show that the implicit bias is approximated (but not exactly equal to) the minimum-norm interpolation in high dimensions, which arises from training on the squared loss. In contrast to prior work, which was tailored to exponentially-tailed losses and used the intermediate support-vector-machine formulation, our framework directly builds on the primal-dual analysis of Ji and Telgarsky (2021), allowing us to provide new approximate equivalences for general convex losses through a novel sensitivity analysis. Our framework also recovers existing exact equivalence results for exponentially-tailed losses across binary and multiclass settings. Finally, we provide evidence for the tightness of our techniques and use our results to demonstrate the effect of certain loss functions designed for out-of-distribution problems on the closed-form solution.

Classification models are expected to perform equally well for different classes, yet in practice, there are often large gaps in their performance. This issue of class bias is widely studied in cases of datasets with sample imbalance, but is relatively overlooked in balanced datasets. In this work, we introduce the concept of spectral imbalance in features as a potential source for class disparities and study the connections between spectral imbalance and class bias in both theory and practice. To build the connection between spectral imbalance and class gap, we develop a theoretical framework for studying class disparities and derive exact expressions for the per-class error in a high-dimensional mixture model setting. We then study this phenomenon in 11 different state-of-the-art pre-trained encoders, and show how our proposed framework can be used to compare the quality of encoders, as well as evaluate and combine data augmentation strategies to mitigate the issue. Our work sheds light on the class-dependent effects of learning, and provides new insights into how state-of-the-art pre-trained features may have unknown biases that can be diagnosed through their spectra.

Overparameterized models that achieve zero training error are observed to generalize well on average, but degrade in performance when faced with data that is under-represented in the training sample. In this work, we study an overparameterized Gaussian mixture model imbued with a spurious feature, and sharply analyze the in-distribution and out-of-distribution test error of a cost-sensitive interpolating solution that incorporates “importance weights”. Compared to recent work [1], [2], our analysis is sharp with matching upper and lower bounds, and significantly weakens required assumptions on data dimensionality. Our error characterizations apply to any choice of importance weights and unveil a novel tradeoff between worst-case robustness to distribution shift and average accuracy as a function of the importance weight magnitude11An extended version of this paper is available at https://arxiv.org/abs/2405.06546..

The classical iteratively reweighted least-squares (IRLS) algorithm aims to recover an unknown signal from linear measurements by performing a sequence of weighted least squares problems, where the weights are recursively updated at each step. Varieties of this algorithm have been shown to achieve favorable empirical performance and theoretical guarantees for sparse recovery and $\ell_p$-norm minimization. Recently, some preliminary connections have also been made between IRLS and certain types of non-convex linear neural network architectures that are observed to exploit low-dimensional structure in high-dimensional linear models. In this work, we provide a unified asymptotic analysis for a family of algorithms that encompasses IRLS, the recently proposed lin-RFM algorithm (which was motivated by feature learning in neural networks), and the alternating minimization algorithm on linear diagonal neural networks. Our analysis operates in a "batched" setting with i.i.d. Gaussian covariates and shows that, with appropriately chosen reweighting policy, the algorithm can achieve favorable performance in only a handful of iterations. We also extend our results to the case of group-sparse recovery and show that leveraging this structure in the reweighting scheme provably improves test error compared to coordinate-wise reweighting.

Modern machine learning models are prone to over-reliance on spurious correlations, which can often lead to poor performance on minority groups. In this paper, we identify surprising and nuanced behavior of finetuned models on worst-group accuracy via comprehensive experiments on four well-established benchmarks across vision and language tasks. We first show that the commonly used class-balancing techniques of mini-batch upsampling and loss upweighting can induce a decrease in worst-group accuracy (WGA) with training epochs, leading to performance no better than without class-balancing. While in some scenarios, removing data to create a class-balanced subset is more effective, we show this depends on group structure and propose a mixture method which can outperform both techniques. Next, we show that scaling pretrained models is generally beneficial for worst-group accuracy, but only in conjunction with appropriate class-balancing. Finally, we identify spectral imbalance in finetuning features as a potential source of group disparities --- minority group covariance matrices incur a larger spectral norm than majority groups once conditioned on the classes. Our results show more nuanced interactions of modern finetuned models with group robustness than was previously known. Our code is available at https://github.com/tmlabonte/revisiting-finetuning.

The paradigm of inverse reinforcement learning (IRL) is used to specify the reward function of an agent purely from its actions and is critical for value alignment and AI safety. While IRL is successful in practice, theoretical guarantees remain nascent. Motivated by the need for IRL in large action spaces with limited data, we consider as a first step the problem of learning from a single sequence of actions (i.e., a demonstration) of a stochastic linear bandit algorithm. When the demonstrator employs the Phased Elimination algorithm, we develop a simple inverse learning procedure that estimates the linear reward function consistently in the time horizon with just a single demonstration. In particular, we show that our inverse learner approximates the true reward parameter within a error of $\mathcal{O}(T^{-\frac{\omega - 1}{2\omega }})$ (where $T$ is the length of the demonstrator’s trajectory and $\omega$ is a constant that depends on the geometry of the action set). We complement this result with an information-theoretic lower bound for any inverse learning procedure. We corroborate our theoretical results with simulations on synthetic data and a demonstration constructed from the MovieLens dataset.

Data augmentation (DA) is a powerful workhorse for bolstering performance in modern machine learning. Specific augmentations like translations and scaling in computer vision are traditionally believed to improve generalization by generating new (artificial) data from the same distribution. However, this traditional viewpoint does not explain the success of prevalent augmentations in modern machine learning (e.g. randomized masking, cutout, mixup), that greatly alter the training data distribution. In this work, we develop a new theoretical framework to characterize the impact of a general class of DA on underparameterized and overparameterized linear model generalization. Our framework reveals that DA induces implicit spectral regularization through a combination of two distinct effects: a) manipulating the relative proportion of eigenvalues of the data covariance matrix in a training-data-dependent manner, and b) uniformly boosting the entire spectrum of the data covariance matrix through ridge regression. These effects, when applied to popular augmentations, give rise to a wide variety of phenomena, including discrepancies in generalization between overparameterized and underparameterized regimes and differences between regression and classification tasks. Our framework highlights the nuanced and sometimes surprising impacts of DA on generalization, and serves as a testbed for novel augmentation design.

We study the problem of estimating the stationary mass---also called the unigram mass---that is missing from a single trajectory of a discrete-time, ergodic Markov chain. This problem has several applications---for example, estimating the stationary missing mass is critical for accurately smoothing probability estimates in sequence models. While the classical Good--Turing estimator from the 1950s has appealing properties for i.i.d. data, it is known to be biased in the Markovian setting, and other heuristic estimators do not come equipped with guarantees. Operating in the general setting in which the size of the state space may be much larger than the length $n$ of the trajectory, we develop a linear-runtime estimator called Windowed Good--Turing (WingIt) and show that its risk decays as $\widetilde{O}(\mathsf{T_{mix}}/n)$, where $\mathsf{T_{mix}}$ denotes the mixing time of the chain in total variation distance. Notably, this rate is independent of the size of the state space and minimax-optimal up to a logarithmic factor in $n / \mathsf{T_{mix}}$. We also present an upper bound on the variance of the missing mass random variable, which may be of independent interest. We extend our estimator to approximate the stationary mass placed on elements occurring with small frequency in the trajectory. Finally, we demonstrate the efficacy of our estimators both in simulations on canonical chains and on sequences constructed from natural language text.

We study the complexity of computing stationary Nash equilibrium (NE) in n-player infinite-horizon general-sum stochastic games. We focus on the problem of computing NE in such stochastic games when each player is restricted to choosing a stationary policy and rewards are discounted. First, we prove that computing such NE is in PPAD (in addition to clearly being PPAD-hard). Second, we consider turn-based specializations of such games where at each state there is at most a single player that can take actions and show that these (seemingly-simpler) games remain PPAD-hard. Third, we show that under further structural assumptions on the rewards computing NE in such turn-based games is possible in polynomial time. Towards achieving these results we establish structural facts about stochastic games of broader utility, including monotonicity of utilities under single-state single-action changes and reductions to settings where each player controls a single state.

Empirical risk minimization (ERM) of neural networks is prone to over-reliance on spurious correlations and poor generalization on minority groups. The recent deep feature reweighting (DFR) technique achieves state-of-the-art group robustness via simple last-layer retraining, but it requires held-out group and class annotations to construct a group-balanced reweighting dataset. In this work, we examine this impractical requirement and find that last-layer retraining can be surprisingly effective with no group annotations (other than for model selection) and only a handful of class annotations. We first show that last-layer retraining can greatly improve worst-group accuracy even when the reweighting dataset has only a small proportion of worst-group data. This implies a "free lunch" where holding out a subset of training data to retrain the last layer can substantially outperform ERM on the entire dataset with no additional data, annotations, or computation for training. To further improve group robustness, we introduce a lightweight method called selective last-layer finetuning (SELF), which constructs the reweighting dataset using misclassifications or disagreements. Our experiments present the first evidence that model disagreement upsamples worst-group data, enabling SELF to nearly match DFR on four well-established benchmarks across vision and language tasks with no group annotations and less than 3% of the held-out class annotations.

First-order optimization methods tend to inherently favor certain solutions over others when minimizing a given training objective with multiple local optima. This phenomenon, known as \emph{implicit bias}, plays a critical role in understanding the generalization capabilities of optimization algorithms. Recent research has revealed that gradient-descent-based methods exhibit an implicit bias for the $\ell_2$-maximal margin classifier in the context of separable binary classification. In contrast, generic optimization methods, such as mirror descent and steepest descent, have been shown to converge to maximal margin classifiers defined by alternative geometries. However, while gradient-descent-based algorithms demonstrate fast implicit bias rates, the implicit bias rates of generic optimization methods have been relatively slow. To address this limitation, in this paper, we present a series of state-of-the-art implicit bias rates for mirror descent and steepest descent algorithms. Our primary technique involves transforming a generic optimization algorithm into an online learning dynamic that solves a regularized bilinear game, providing a unified framework for analyzing the implicit bias of various optimization methods. The accelerated rates are derived leveraging the regret bounds of online learning algorithms within this game framework.

The literature on “benign overfitting” in overparameterized models has been mostly restricted to regression or binary classification; however, modern machine learning operates in the multiclass setting. Motivated by this discrepancy, we study benign overfitting in multiclass linear classification. Specifically, we consider the following training algorithms on separable data: 1) empirical risk minimization (ERM) with cross-entropy loss, which converges to the multiclass support vector machine (SVM) solution; 2) ERM with least-squares loss, which converges to the min-norm interpolating (MNI) solution; and 3) the one-vs-all SVM classifier. First, we provide a simple sufficient deterministic condition under which all three algorithms lead to classifiers that interpolate the training data and have equal accuracy. When the data is generated from Gaussian mixtures or a multinomial logistic model, this condition holds under high enough effective overparameterization. We also show that this sufficient condition is satisfied under “neural collapse”, a phenomenon that is observed in training deep neural networks. Second, we derive novel bounds on the accuracy of the MNI classifier, thereby showing that all three training algorithms lead to benign overfitting under sufficient overparameterization. Ultimately, our analysis shows that good generalization is possible for SVM solutions beyond the realm in which typical margin-based bounds apply.

We introduce the “inverse bandit” problem of estimating the rewards of a multi-armed bandit instance from observing the learning process of a low-regret demonstrator. Existing approaches to the related problem of inverse reinforcement learning assume the execution of an optimal policy, and thereby suffer from an identifiability issue. In contrast, we propose to leverage the demonstrator’s behavior en route to optimality, and in particular, the exploration phase, for reward estimation. We begin by establishing a general information-theoretic lower bound under this paradigm that applies to any demonstrator algorithm, which characterizes a fundamental tradeoff between reward estimation and the amount of exploration of the demonstrator. Then, we develop simple and efficient reward estimators for upper-confidence-based demonstrator algorithms that attain the optimal tradeoff, showing in particular that consistent reward estimation—free of identifiability issues—is possible under our paradigm. Extensive simulations on both synthetic and semi-synthetic data corroborate our theoretical results.

Overparametrized neural networks tend to perfectly fit noisy training data yet generalize well on test data. Inspired by this empirical observation, recent work has sought to understand this phenomenon of benign overfitting or harmless interpolation in the much simpler linear model. Previous theoretical work critically assumes that either the data features are statistically independent or the input data is high-dimensional; this precludes general nonparametric settings with structured feature maps. In this paper, we present a general and flexible framework for upper bounding regression and classification risk in a reproducing kernel Hilbert space. A key contribution is that our framework describes precise sufficient conditions on the data Gram matrix under which harmless interpolation occurs. Our results recover prior independent-features results (with a much simpler analysis), but they furthermore show that harmless interpolation can occur in more general settings such as features that are a bounded orthonormal system. Furthermore, our results show an asymptotic separation between classification and regression performance in a manner that was previously only shown for Gaussian features.

Model selection in contextual bandits is an important complementary problem to regret minimization with respect to a fixed model class. We consider the simplest non-trivial instance of model-selection: distinguishing a simple multi-armed bandit problem from a linear contextual bandit problem. Even in this instance, current state-of-the-art methods explore in a suboptimal manner and require strong "feature-diversity" conditions. In this paper, we introduce new algorithms that a) explore in a data-adaptive manner, and b) provide model selection guarantees of the form O(d^{\alpha} T^{1 - \alpha}) with no feature diversity conditions whatsoever, where d denotes the dimension of the linear model and T denotes the total number of rounds. The first algorithm enjoys a "best-of-both-worlds" property, recovering two prior results that hold under distinct distributional assumptions, simultaneously. The second removes distributional assumptions altogether, expanding the scope for tractable model selection. Our approach extends to model selection among nested linear contextual bandits under some additional assumptions.

The classical algorithms for online learning and decision-making have the benefit of achieving the optimal performance guarantees, but suffer from computational complexity limitations when implemented at scale. More recent sophisticated techniques, which we refer to as $\textit{oracle-efficient}$ methods, address this problem by dispatching to an $\textit{offline optimization oracle}$ that can search through an exponentially-large (or even infinite) space of decisions and select that which performed the best on any dataset. But despite the benefits of computational feasibility, most oracle-efficient algorithms exhibit one major limitation: while performing well in worst-case settings, they do not adapt well to friendly environments. In this paper we consider two such friendly scenarios, (a) "small-loss" problems and (b) IID data. We provide a new framework for designing follow-the-perturbed-leader algorithms that are oracle-efficient and adapt well to the small-loss environment, under a particular condition which we call $\textit{approximability}$ (which is spiritually related to sufficient conditions provided in (Dudík et al., 2020)). We identify a series of real-world settings, including online auctions and transductive online classification, for which approximability holds. We also extend the algorithm to an IID data setting and establish a "best-of-both-worlds" bound in the oracle-efficient setting.

The support vector machine (SVM) is a well-established classification method whose name refers to the particular training examples, called support vectors, that determine the maximum margin separating hyperplane. The SVM classifier is known to enjoy good generalization properties when the number of support vectors is small compared to the number of training examples. However, recent research has shown that in sufficiently high-dimensional linear classification problems, the SVM can generalize well despite a proliferation of support vectors where all training examples are support vectors. In this paper, we identify new deterministic equivalences for this phenomenon of support vector proliferation, and use them to (1) substantially broaden the conditions under which the phenomenon occurs in high-dimensional settings, and (2) prove a nearly matching converse result.

Deep reinforcement learning has achieved impressive successes yet often requires a very large amount of interaction data. This result is perhaps unsurprising, as using complicated function approximation often requires more data to fit, and early theoretical results on linear Markov decision processes provide regret bounds that scale with the dimension of the linear approximation. Ideally, we would like to automatically identify the minimal dimension of the approximation that is sufficient to encode an optimal policy. Towards this end, we consider the problem of model selection in RL with function approximation, given a set of candidate RL algorithms with known regret guarantees. The learner’s goal is to adapt to the complexity of the optimal algorithm without knowing it a priori. We present a meta-algorithm that successively rejects increasingly complex models using a simple statistical test. Given at least one candidate that satisfies realizability, we prove the meta-algorithm adapts to the optimal complexity with regret that is only marginally suboptimal in the number of episodes and number of candidate algorithms. The dimension and horizon dependencies remain optimal with respect to the best candidate, and our meta-algorithmic approach is flexible to incorporate multiple candidate algorithms and models. Finally, we show that the meta-algorithm automatically admits significantly improved instance-dependent regret bounds that depend on the gaps between the maximal values attainable by the candidates.

The growing literature on "benign overfitting" in overparameterized models has been mostly restricted to regression or binary classification settings; however, most success stories of modern machine learning have been recorded in multiclass settings. Motivated by this discrepancy, we study benign overfitting in multiclass linear classification. Specifically, we consider the following popular training algorithms on separable data: (i) empirical risk minimization (ERM) with cross-entropy loss, which converges to the multiclass support vector machine (SVM) solution; (ii) ERM with least-squares loss, which converges to the min-norm interpolating (MNI) solution; and, (iii) the one-vs-all SVM classifier. Our first key finding is that under a simple sufficient condition, all three algorithms lead to classifiers that interpolate the training data and have equal accuracy. When the data is generated from Gaussian mixtures or a multinomial logistic model, this condition holds under high enough effective overparameterization. Second, we derive novel error bounds on the accuracy of the MNI classifier, thereby showing that all three training algorithms lead to benign overfitting under sufficient overparameterization. Ultimately, our analysis shows that good generalization is possible for SVM solutions beyond the realm in which typical margin-based bounds apply.

We compare classification and regression tasks in an overparameterized linear model with Gaussian features. On the one hand, we show that with sufficient overparameterization all training points are support vectors: solutions obtained by least-squares minimum-norm interpolation, typically used for regression, are identical to those produced by the hard-margin support vector machine (SVM) that minimizes the hinge loss, typically used for training classifiers. On the other hand, we show that there exist regimes where these interpolating solutions generalize well when evaluated by the 0-1 test loss function, but do not generalize if evaluated by the square loss function, i.e. they approach the null risk. Our results demonstrate the very different roles and properties of loss functions used at the training phase (optimization) and the testing phase (generalization).

We consider the stochastic linear (multi-armed) contextual bandit problem with the possibility of hidden simple multi-armed bandit structure in which the rewards are independent of the contextual information. Algorithms that are designed solely for one of the regimes are known to be sub-optimal for their alternate regime. We design a single computationally efficient algorithm that simultaneously obtains problem-dependent optimal regret rates in the simple multi-armed bandit regime and minimax optimal regret rates in the linear contextual bandit regime, without knowing a priori which of the two models generates the rewards. These results are proved under the condition of stochasticity of contextual information over multiple rounds. Our results should be viewed as a step towards principled data-dependent policy class selection for contextual bandits.

We introduce algorithms for online, full-information prediction that are computationally efficient and competitive with contextual tree experts of unknown complexity, in both probabilistic and adversarial settings. We incorporate a novel probabilistic framework of structural risk minimization into existing adaptive algorithms and show that we can robustly learn not only the presence of stochastic structure when it exists, but also the correct model order. When the stochastic data is actually realized from a predictor in the model class considered, we obtain regret bounds that are competitive with the regret of an optimal algorithm that possesses strong side information about both the true model order and whether the process generating the data is stochastic or adversarial. In cases where the data does not arise from any of the models, our algorithm selects models of higher order as we play more rounds. We display empirically improved \textit{overall prediction error} over other adversarially robust approaches.

A continuing mystery in understanding the empirical success of deep neural networks has been in their ability to achieve zero training error and yet generalize well, even when the training data is noisy and there are more parameters than data points. We investigate this "overparametrization" phenomena in the classical underdetermined linear regression problem, where all solutions that minimize training error interpolate the data, including noise. We give a bound on how well such interpolative solutions can generalize to fresh test data, and show that this bound generically decays to zero with the number of extra features, thus characterizing an explicit benefit of overparameterization. For appropriately sparse linear models, we provide a hybrid interpolating scheme (combining classical sparse recovery schemes with harmless noise-fitting) to achieve generalization error close to the bound on interpolative solutions.

Recent advances in dynamic spectrum sharing have led to renewed focus on the structure of regulatory games between a primary user and a secondary user of a spectrum band. The primary user has to decide to what extent it invokes the services of the regulator, and the secondary user has to decide how to operate in the spectrum band. This paper builds on a mathematical model for light-handed regulation using “spectrum jails” to show that the order of play and amount of commitment matters. A primary that can commit to its strategy before the game is able to increase its equilibrium payoff, even when the secondary best responds to the committed strategy. We compare the ensuing Stackelberg game with the simultaneous primary-secondary game. We also introduce a new concept of partial commitment by which the primary can only commit to a range of strategies using a finite number of bits. We show explicitly that the more the primary commits to, the more it benefits, and that Stackelberg commitment can be understood as a limit of infinite “commitment bits”.

Spectrum has traditionally been allocated for single uses and by now most of the “prime” spectrum has well-entrenched incumbent users. When a new service needs spectrum, there are two qualitatively distinct ways of making bandwidth available for it. A swath of incumbent users can be removed from a band, with the cleared band being reallocated for the new service. Alternatively, the new users can be allowed to utilize the interstitial spectrum holes (i.e. whitespaces) between incumbent users, with the requirement to protect the incumbents' QoS. But these can also be used in combination by partially clearing a band and opening up the rest for whitespace-style sharing. In this case, the ability of regulators to “repack” incumbents, e.g. alter their operating channels, can reduce the need to evict them. An open question has been how whitespaces and partial spectrum clearing interact with each other and the ability to repack incumbents. Do efficient repacks completely eliminate whitespaces? The USA FCC's upcoming incentive auction in the TV bands is the first large-scale attempt to repack a major band of spectrum in order to clear spectrum for LTE. This auction is meant to navigate the tradeoff between incumbent TV services and LTE networks. In preparation, the FCC has made a large and complex data set of repacking constraints available for the first time. We have repurposed this data and built our own repacking engine in order to study a more general version of the tradeoff between whitespaces and cleared spectrum. We conclude that (1) repacking enables clearing of significantly more spectrum than just removing incumbents; (2) the total amount of spectrum available for new uses is relatively insensitive to how incumbents are removed; (3) efficient repackings basically trade whitespace spectrum for cleared spectrum; (4) even the most efficient repackings leave plenty of whitespace - an amount that can be comparable with the amount of cleared spectrum.

We explore the problem of deterministically constructing frames and matrices with low coherence, which arises in areas such as compressive sensing, spherical codes, and MIMO communications. In particular, we present a generalization of the familiar harmonic frame by selecting a subset of rows of the generalized discrete Fourier transform matrix over finite groups. We apply our methods to the group SL2(Fq) and show how to produce frames with remarkably low coherence, for which we provide upper bounds.