Knowledge distillation deploys complex machine learning models in resource-constrained environments by training a smaller student model to emulate internal representations of a complex teacher model. However, the teacher’s representations can also encode nuisance or additional information not relevant to the downstream task. Distilling such irrelevant information can actually impede the performance of a capacity-limited student model. This observation motivates our primary question: What are the information-theoretic limits of knowledge distillation? To this end, we leverage Partial Information Decomposition to quantify and explain the transferred knowledge and knowledge left to distill for a downstream task. We theoretically demonstrate that the task-relevant transferred knowledge is succinctly captured by the measure of redundant information about the task between the teacher and student. We propose a novel multi-level optimization to incorporate redundant information as a regularizer, leading to our framework of Redundant Information Distillation (RID). RID leads to more resilient and effective distillation under nuisance teachers as it succinctly quantifies task-relevant knowledge rather than simply aligning student and teacher representations.

We study beyond worst-case dimensionality reduction for $s$-sparse vectors (vectors with at most $s$ non-zero coordinates). Our work is divided into two parts, each focusing on a different facet of beyond worst-case analysis: \noindent (a) We first consider average-case guarantees for embedding $s$-sparse vectors. Here, a well-known folklore upper bound based on the birthday-paradox states: For any collection $X$ of $s$-sparse vectors in $\mathbb{R}^d$, there exists a linear map $A: \mathbb{R}^d \rightarrow \mathbb{R}^{O(s^2)}$ which \emph{exactly} preserves the norm of $99\%$ of the vectors in $X$ in any $\ell_p$ norm (as opposed to the usual setting where guarantees hold for all vectors). We provide novel lower bounds showing that this is indeed optimal in many settings. Specifically, any oblivious linear map satisfying similar average-case guarantees must map to $\Omega(s^2)$ dimensions. The same lower bound also holds for a wider class of sufficiently smooth maps, including `encoder-decoder schemes', where we compare the norm of the original vector to that of a smooth function of the embedding. These lower bounds reveal a surprising separation result for smooth embeddings of sparse vectors, as an upper bound of $O(s \log(d))$ is possible if we instead use arbitrary functions, e.g., via compressed sensing algorithms. (b) Given these lower bounds, we specialize to sparse \emph{non-negative} vectors to hopes of improved upper bounds. For a dataset $X$ of non-negative $s$-sparse vectors and any $p \ge 1$, we can non-linearly embed $X$ to $O(s\log(|X|s)/\varepsilon^2)$ dimensions while preserving all pairwise distances in $\ell_p$ norm up to $1\pm \varepsilon$, with no dependence on $p$. Surprisingly, the non-negativity assumption enables much smaller embeddings than arbitrary sparse vectors, where the best known bound suffers an exponential $(\log |X|)^{O(p)}$ dependence. Our map also guarantees \emph{exact} dimensionality reduction for the $\ell_{\infty}$ norm by embedding $X$ into $O(s\log |X|)$ dimensions, which is tight. We further give separation results showing that both the non-linearity of $f$ and the non-negativity of $X$ are necessary, and provide downstream algorithmic improvements using our embedding.

AI applications across classification, fairness, and human interaction often implicitly require ontologies of social concepts. Constructing these well – especially when there are many relevant categories – is a controversial task but is crucial for achieving meaningful inclusivity. Here, we focus on developing a pragmatic ontology of belief systems, which isa complex and often controversial space. By iterating on our community-based design until mutual agreement is reached, we found that epistemological methods were best for categorizing the fundamental ways beliefs differ, maximally respecting our principles of inclusivity and brevity. We demonstrate our methodology’s utility and interpretability via user studies in term annotation and sentiment analysis experiments for belief fairness in language models

Fast changing states or volatile environments pose a significant challenge to online optimization, which needs to perform rapid adaptation under limited observation. In this paper, we give query and regret optimal bandit algorithms under the strict notion of strongly adaptive regret, which measures the maximum regret over any contiguous interval $I$. Due to its worst-case nature, there is an almost-linear $\Omega(|I|^{1-\epsilon})$ regret lower bound, when only one query per round is allowed [Daniely el al, ICML 2015]. Surprisingly, with just two queries per round, we give Strongly Adaptive Bandit Learner (StABL) that achieves $\widetilde{O}(\sqrt{n|I|})$ adaptive regret for multi-armed bandits with $n$ arms. The bound is tight and cannot be improved in general. Our algorithm leverages a multiplicative update scheme of varying stepsizes and a carefully chosen observation distribution to control the variance. Furthermore, we extend our results and provide optimal algorithms in the bandit convex optimization setting. Finally, we empirically demonstrate the superior performance of our algorithms under volatile environments and for downstream tasks, such as algorithm selection for hyperparameter optimization.

Model overconfidence and poor calibration are common in machine learning and difficult to account for when applying standard empirical risk minimization. In this work, we propose a novel method to alleviate these problems that we call odd-$k$-out learning (OKO), which minimizes the cross-entropy error for sets rather than for single examples. This naturally allows the model to capture correlations across data examples and achieves both better accuracy and calibration, especially in limited training data and class-imbalanced regimes. Perhaps surprisingly, OKO often yields better calibration even when training with hard labels and dropping any additional calibration parameter tuning, such as temperature scaling. We demonstrate this in extensive experimental analyses and provide a mathematical theory to interpret our findings. We emphasize that OKO is a general framework that can be easily adapted to many settings and a trained model can be applied to single examples at inference time, without significant run-time overhead or architecture changes.

Preference learning algorithms (e.g., RLHF and DPO) are frequently used to steer LLMs to produce generations that are more preferred by humans, but our understanding of their inner workings is still limited. In this work, we study the conventional wisdom that preference learning trains models to assign higher likelihoods to more preferred outputs than less preferred outputs, measured via *ranking accuracy*. Surprisingly, we find that most state-of-the-art preference-tuned models achieve a ranking accuracy of less than 60% on common preference datasets. We furthermore derive the *idealized ranking accuracy* that a preference-tuned LLM would achieve if it optimized the DPO or RLHF objective perfectly. We demonstrate that existing models exhibit a significant *alignment gap* -- *i.e.*, a gap between the observed and idealized ranking accuracies. We attribute this discrepancy to the DPO objective, which is empirically and theoretically ill-suited to correct even mild ranking errors in the reference model, and derive a simple and efficient formula for quantifying the difficulty of learning a given preference datapoint. Finally, we demonstrate that ranking accuracy strongly correlates with the empirically popular win rate metric when the model is close to the reference model used in the objective, shedding further light on the differences between on-policy (e.g., RLHF) and off-policy (e.g., DPO) preference learning algorithms.

Scalarization is a general, parallizable technique that can be deployed in any multiobjective setting to reduce multiple objectives into one, yet some have dismissed this versatile approach because linear scalarizations cannot explore concave regions of the Pareto frontier. To that end, we aim to find simple non-linear scalarizations that provably explore a diverse set of $k$ objectives on the Pareto frontier, as measured by the dominated hypervolume. We show that hypervolume scalarizations with uniformly random weights achieves an optimal sublinear hypervolume regret bound of $O(T^{-1/k})$, with matching lower bounds that preclude any algorithm from doing better asymptotically. For the setting of multiobjective stochastic linear bandits, we utilize properties of hypervolume scalarizations to derive a novel non-Euclidean analysis to get regret bounds of $\tilde{O}( d T^{-1/2} + T^{-1/k})$, removing unnecessary $\text{poly}(k)$ dependencies. We support our theory with strong empirical performance of using non-linear scalarizations that outperforms both their linear counterparts and other standard multiobjective algorithms in a variety of natural settings.

Inspired by fast algorithms in natural language processing, we study low rank approximation in the entrywise transformed setting where we want to find a good rank $k$ approximation to $f(U \cdot V)$, where $U, V^\top \in \mathbb{R}^{n \times r}$ are given, $r = O(\log(n))$, and $f(x)$ is a general scalar function. Previous work in sublinear low rank approximation has shown that if both (1) $U = V^\top$ and (2) $f(x)$ is a PSD kernel function, then there is an $O(nk^{\omega-1})$ time constant relative error approximation algorithm, where $\omega \approx 2.376$ is the exponent of matrix multiplication. We give the first conditional time hardness results for this problem, demonstrating that both conditions (1) and (2) are in fact necessary for getting better than $n^{2-o(1)}$ time for a relative error low rank approximation for a wide class of functions. We give novel reductions from the Strong Exponential Time Hypothesis (SETH) that rely on lower bounding the leverage scores of flat sparse vectors and hold even when the rank of the transformed matrix $f(UV)$ and the target rank are $n^{o(1)}$, and when $U = V^\top$. Furthermore, even when $f(x) = x^p$ is a simple polynomial, we give runtime lower bounds in the case when $U \neq V^\top$ of the form $\Omega(\min(n^{2-o(1)}, \Omega(2^p)))$. Lastly, we demonstrate that our lower bounds are tight by giving an $O(n \cdot \text{poly}(k, 2^p, 1/\epsilon))$ time relative error approximation algorithm and a fast $O(n \cdot \text{poly}(k, p, 1/\epsilon))$ additive error approximation using fast tensor-based sketching. Additionally, since our low rank algorithms rely on matrix-vector product subroutines, our lower bounds extend to show that computing $f(UV)W$, for even a small matrix $W$, requires $\Omega(n^{2-o(1)})$ time.

We study the setting of optimizing with bandit feedback with additional prior knowledge provided to the learner in the form of an initial hint of the optimal action. We present a novel algorithm for stochastic linear bandits that uses this hint to improve its regret to $\tilde O(\sqrt{T})$ when the hint is accurate, while maintaining a minimax-optimal $\tilde O(d\sqrt{T})$ regret independent of the quality of the hint. Furthermore, we provide a Pareto frontier of tight tradeoffs between best-case and worst-case regret, with matching lower bounds. Perhaps surprisingly, our work shows that leveraging a hint shows provable gains without sacrificing worst-case performance, implying that our algorithm adapts to the quality of the hint for free. We also provide an extension of our algorithm to the case of $m$ initial hints, showing that we can achieve a $\tilde O(m^{2/3}\sqrt{T})$ regret.

Meta-learning hyperparameter optimization (HPO) algorithms from prior experiments is a promising approach to improve optimization efficiency over objective functions from a similar distribution. However, existing methods are restricted to learning from experiments sharing the same set of hyperparameters. In this paper, we introduce the OptFormer, the first text-based Transformer HPO framework that provides a universal end-to-end interface for jointly learning policy and function prediction when trained on vast tuning data from the wild, such as Google’s Vizier database, one of the world’s largest HPO datasets. Our extensive experiments demonstrate that the OptFormer can simultaneously imitate at least 7 different HPO algorithms, which can be further improved via its function uncertainty estimates. Compared to a Gaussian Process, the OptFormer also learns a robust prior distribution for hyperparameter response functions, and can thereby provide more accurate and better calibrated predictions. This work paves the path to future extensions for training a Transformer-based model as a general HPO optimizer.

We consider the problem of minimizing the number of matrix-vector queries needed for accurate trace estimation in the dynamic setting where our underlying matrix is changing slowly, such as during an optimization process. Specifically, for any $m$ matrices $\mathbf{A}_1,...,\mathbf{A}_m$ with consecutive differences bounded in Schatten-$1$ norm by $\alpha$, we provide a novel binary tree summation procedure that simultaneously estimates all $m$ traces up to $\epsilon$ error with $\delta$ failure probability with an optimal query complexity of $\widetilde{O}(m \alpha\sqrt{\log(1/\delta)}/\epsilon + m\log(1/\delta))$, improving the dependence on both $\alpha$ and $\delta$ from Dharangutte and Musco (NeurIPS, 2021). Our procedure works without additional norm bounds on $\mathbf{A}_i$ and can be generalized to a bound for the $p$-th Schatten norm for $p \in [1,2]$, giving a complexity of $\widetilde{O}(m \alpha(\sqrt{\log(1/\delta)}/\epsilon)^p +m \log(1/\delta))$. By using novel reductions to communication complexity and information-theoretic analyses of Gaussian matrices, we provide matching lower bounds for static and dynamic trace estimation in all relevant parameters, including the failure probability. Our lower bounds (1) give the first tight bounds for Hutchinson's estimator in the matrix-vector product model with Frobenius norm error {\it even in the static setting}, and (2) are the first unconditional lower bounds for dynamic trace estimation, resolving open questions of prior work.

Matrix trace estimation is ubiquitous in machine learning applications and has traditionally relied on Hutchinson's method, which requires $O(\log(1/\delta)/\epsilon^2)$ matrix-vector product queries to achieve a $(1 \pm \epsilon)$-multiplicative approximation to $\text{trace}(A)$ with failure probability $\delta$ on positive-semidefinite input matrices $A$. Recently, the Hutch++ algorithm was proposed, which reduces the number of matrix-vector queries from $O(1/\epsilon^2)$ to the optimal $O(1/\epsilon)$, and the algorithm succeeds with constant probability. However, in the high probability setting, the non-adaptive Hutch++ algorithm suffers an extra $O(\sqrt{\log(1/\delta)})$ multiplicative factor in its query complexity. Non-adaptive methods are important, as they correspond to sketching algorithms, which are mergeable, highly parallelizable, and provide low-memory streaming algorithms as well as low-communication distributed protocols. In this work, we close the gap between non-adaptive and adaptive algorithms, showing that even non-adaptive algorithms can achieve $O(\sqrt{\log(1/\delta)}/\epsilon + \log(1/\delta))$ matrix-vector products. In addition, we prove matching lower bounds demonstrating that, up to a $\log \log(1/\delta)$ factor, no further improvement in the dependence on $\delta$ or $\epsilon$ is possible by any non-adaptive algorithm. Finally, our experiments demonstrate the superior performance of our sketch over the adaptive Hutch++ algorithm, which is less parallelizable, as well as over the non-adaptive Hutchinson's method.

Incremental tree building (INC) is a new phylogeny estimation method that has been proven to be absolute fast converging under standard sequence evolution models. A variant of INC, called Constrained-INC, is designed for use in divide-and-conquer pipelines for phylogeny estimation where a set of species is divided into disjoint subsets, trees are computed on the subsets using a selected base method, and then the subset trees are combined together. We evaluate the accuracy of INC and Constrained-INC for gene tree and species tree estimation on simulated datasets, and compare it to similar pipelines using NJMerge (another method that merges disjoint trees). For gene tree estimation, we find that INC has very poor accuracy in comparison to standard methods, and even Constrained-INC(using maximum likelihood methods to compute constraint trees) does not match the accuracy of the better maximum likelihood methods. Results for species trees are somewhat different, with Constrained-INC coming close to the accuracy of the best species tree estimation methods, while being much faster; furthermore, using Constrained-INC allows species tree estimation methods to scale to large datasets within limited computational resources. Overall, this study exposes the benefits and limitations of divide-and-conquer strategies for large-scale phylogenetic tree estimation.

Single-objective black box optimization (also known as zeroth-order optimization) is the process of minimizing a scalar objective $f(x)$, given evaluations at adaptively chosen inputs $x$. In this paper, we consider multi-objective optimization, where $f(x)$ outputs a vector of possibly competing objectives and the goal is to converge to the Pareto frontier. Quantitatively, we wish to maximize the standard \emph{hypervolume indicator} metric, which measures the dominated hypervolume of the entire set of chosen inputs. In this paper, we introduce a novel scalarization function, which we term the \emph{hypervolume scalarization}, and show that drawing random scalarizations from an appropriately chosen distribution can be used to efficiently approximate the \emph{hypervolume indicator} metric. We utilize this connection to show that Bayesian optimization with our scalarization via common acquisition functions, such as Thompson Sampling or Upper Confidence Bound, provably converges to the whole Pareto frontier by deriving tight \emph{hypervolume regret} bounds on the order of $\widetilde{O}(\sqrt{T})$. Furthermore, we highlight the general utility of our scalarization framework by showing that any provably convergent single-objective optimization process can be converted to a multi-objective optimization process with provable convergence guarantees.

Many convex problems in machine learning and computer science share the same form: \begin{align*} \min_{x} \sum_{i} f_i( A_i x + b_i), \end{align*} where $f_i$ are convex functions on $\R^{n_i}$ with constant $n_i$, $A_i \in \R^{n_i \times d}$, $b_i \in \R^{n_i}$ and $\sum_i n_i = n$. This problem generalizes linear programming and includes many problems in empirical risk minimization. In this paper, we give an algorithm that runs in time \begin{align*} O^* ( ( n^{\omega} + n^{2.5 - \alpha/2} + n^{2+ 1/6} ) \log (n / \delta) ) \end{align*} where $\omega$ is the exponent of matrix multiplication, $\alpha$ is the dual exponent of matrix multiplication, and $\delta$ is the relative accuracy. Note that the runtime has only a log dependence on the condition numbers or other data dependent parameters and these are captured in $\delta$. For the current bound $\omega \sim 2.38$ [Vassilevska Williams’12, Le Gall’14] and $\alpha \sim 0.31$ [Le Gall, Urrutia’18], our runtime $O^* ( n^{\omega} \log (n / \delta))$ matches the current best for solving a dense least squares regression problem, a special case of the problem we consider. Very recently, [Alman’18] proved that all the current known techniques can not give a better $\omega$ below $2.168$ which is larger than our $2+1/6$. Our result generalizes the very recent result of solving linear programs in the current matrix multiplication time [Cohen, Lee, Song’19] to a more broad class of problems. Our algorithm proposes two concepts which are different from [Cohen, Lee, Song’19] :\ $\bullet$ We give a robust deterministic central path method, whereas the previous one is a stochastic central path which updates weights by a random sparse vector. \ $\bullet$ We propose an efficient data-structure to maintain the central path of interior point methods even when the weights update vector is dense.

The classical low rank approximation problem is to find a rank $k$ matrix $UV$ (where $U$ has $k$ columns and $V$ has $k$ rows) that minimizes the Frobenius norm of $A - UV$. Although this problem can be solved efficiently, we study an NP-hard variant of this problem that involves weights and regularization. A previous paper of [Razenshteyn et al. '16] derived a polynomial time algorithm for weighted low rank approximation with constant rank. We derive provably sharper guarantees for the regularized version by obtaining parameterized complexity bounds in terms of the statistical dimension rather than the rank, allowing for a rank-independent runtime that can be significantly faster. Our improvement comes from applying sharper matrix concentration bounds, using a novel conditioning technique, and proving structural theorems for regularized low rank problems.

We consider the phylogenetic tree reconstruction problem with insertions and deletions (indels). Phylogenetic algorithms proceed under a model where sequences evolve down the model tree, and given sequences at the leaves, the problem is to reconstruct the model tree with high probability. Traditionally, sequences mutate by substitution-only processes, although some recent work considers evolutionary processes with insertions and deletions. In this paper, we improve on previous work by giving a reconstruction algorithm that simultaneously has O(poly logn) sequence length and tolerates constant indel probabilities on each edge. Our recursively-reconstructed distance-based technique provably outputs the model tree when the model tree has O(poly logn) diameter and discretized branch lengths, allowing for the probability of insertion and deletion to be non-uniform and asymmetric on each edge. Our polylogarithmic sequence length bounds improve significantly over previous polynomial sequence length bounds and match sequence length bounds in the substitution-only models of phylogenetic evolution, thereby challenging the idea that many global misalignments caused by insertions and deletions when pindel is large are a fundamental obstruction to reconstruction with short sequences.

We study whether a depth two neural network can learn another depth two network using gradient descent. Assuming a linear output node, we show that the question of whether gradient descent converges to the target function is equivalent to the following question in electrodynamics: Given k fixed protons in R^d, and k electrons, each moving due to the attractive force from the protons and repulsive force from the remaining electrons, whether at equilibrium all the electrons will be matched up with the protons, up to a permutation. Under the standard electrical force, this follows from the classic Earnshaw's theorem. In our setting, the force is determined by the activation function and the input distribution. Building on this equivalence, we prove the existence of an activation function such that gradient descent learns at least one of the hidden nodes in the target network. Iterating, we show that gradient descent can be used to learn the entire network one node at a time.

Absolute fast converging (AFC) phylogeny estimation methods are ones that have been proven to recover the true tree with high probability given sequences whose lengths are polynomial in the number of number of leaves in the tree (once the shortest and longest branch lengths are fixed). While there has been a large literature on AFC methods, the best in terms of empirical performance was DCM_NJ, published in SODA 2001. The main empirical advantage of DCM_NJ over other AFC methods is its use of neighbor joining (NJ) to construct trees on smaller taxon subsets, which are then combined into a tree on the full set of species using a supertree method; in contrast, the other AFC methods in essence depend on quartet trees that are computed independently of each other, which reduces accuracy compared to neighbor joining. However, DCM_NJ is unlikely to scale to large datasets due to its reliance on supertree methods, as no current supertree methods are able to scale to large datasets with high accuracy. In this study we present a new approach to large-scale phylogeny estimation that shares some of the features of DCM_NJ but bypasses the use of supertree methods. We prove that this new approach is AFC and uses polynomial time. Furthermore, we describe variations on this basic approach that can be used with leaf-disjoint constraint trees (computed using methods such as maximum likelihood) to produce other AFC methods that are likely to provide even better accuracy. Thus, we present a new generalizable technique for large-scale tree estimation that is designed to improve scalability for phylogeny estimation methods to ultra-large datasets, and that can be used in a variety of settings (including tree estimation from unaligned sequences, and species tree estimation from gene trees).