More Like this

1. A projector-based approach to quantifying total and excess uncertainties for sketched linear regression

   https://doi.org/10.1093/imaiai/iaab016  

   Chi, Jocelyn T; Ipsen, Ilse C (August 2021, Information and Inference: A Journal of the IMA)

   Abstract Linear regression is a classic method of data analysis. In recent years, sketching—a method of dimension reduction using random sampling, random projections or both—has gained popularity as an effective computational approximation when the number of observations greatly exceeds the number of variables. In this paper, we address the following question: how does sketching affect the statistical properties of the solution and key quantities derived from it? To answer this question, we present a projector-based approach to sketched linear regression that is exact and that requires minimal assumptions on the sketching matrix. Therefore, downstream analyses hold exactly and generally for all sketching schemes. Additionally, a projector-based approach enables derivation of key quantities from classic linear regression that account for the combined model- and algorithm-induced uncertainties. We demonstrate the usefulness of a projector-based approach in quantifying and enabling insight on excess uncertainties and bias-variance decompositions for sketched linear regression. Finally, we demonstrate how the insights from our projector-based analyses can be used to produce practical sketching diagnostics to aid the design of judicious sketching schemes. 

   more » « less
2. DiPS: Differentiable Policy for Sketching in Recommender Systems

   https://doi.org/10.1609/aaai.v36i6.20625  

   Ghosh, Aritra; Mitra, Saayan; Lan, Andrew (June 2022, Proceedings of the AAAI Conference on Artificial Intelligence)

   In sequential recommender system applications, it is important to develop models that can capture users' evolving interest over time to successfully recommend future items that they are likely to interact with. For users with long histories, typical models based on recurrent neural networks tend to forget important items in the distant past. Recent works have shown that storing a small sketch of past items can improve sequential recommendation tasks. However, these works all rely on static sketching policies, i.e., heuristics to select items to keep in the sketch, which are not necessarily optimal and cannot improve over time with more training data. In this paper, we propose a differentiable policy for sketching (DiPS), a framework that learns a data-driven sketching policy in an end-to-end manner together with the recommender system model to explicitly maximize recommendation quality in the future. We also propose an approximate estimator of the gradient for optimizing the sketching algorithm parameters that is computationally efficient. We verify the effectiveness of DiPS on real-world datasets under various practical settings and show that it requires up to 50% fewer sketch items to reach the same predictive quality than existing sketching policies. 

   more » « less
3. DiPS: Differentiable Policy for Sketching in Recommender Systems

   Aritra Ghosh, Saayan Mitra (February 2022, AAAI Conference on Artificial Intelligence (AAAI))

   In sequential recommender system applications, it is important to develop models that can capture users’ evolving interest over time to successfully recommend future items that they are likely to interact with. For users with long histories, typical models based on recurrent neural networks tend to forget important items in the distant past. Recent works have shown that storing a small sketch of past items can improve sequential recommendation tasks. However, these works all rely on static sketching policies, i.e., heuristics to select items to keep in the sketch, which are not necessarily optimal and cannot improve over time with more training data. In this paper, we propose a differentiable policy for sketching (DiPS), a framework that learns a data-driven sketching policy in an end-to-end manner together with the recommender system model to explicitly maximize recommendation quality in the future. We also propose an approximate estimator of the gradient for optimizing the sketching algorithm parameters that is computationally efficient. We verify the effectiveness of DiPS on real-world datasets under various practical settings and show that it requires up to 50% fewer sketch items to reach the same predictive quality than existing sketching policies. 

   more » « less
4. Randomized Algorithms for Symmetric Nonnegative Matrix Factorization

   https://doi.org/10.1137/24M1638355  

   Hayashi, Koby; Aksoy, Sinan G; Ballard, Grey; Park, Haesun (March 2025, SIAM Journal on Matrix Analysis and Applications)

   Symmetric Nonnegative Matrix Factorization (SymNMF) is a technique in data analysis and machine learning that approximates a symmetric matrix with a product of a nonnegative, low-rank matrix and its transpose. To design faster and more scalable algorithms for SymNMF, we develop two randomized algorithms for its computation. The first algorithm uses randomized matrix sketching to compute an initial low-rank approximation to the input matrix and proceeds to rapidly compute a SymNMF of the approximation. The second algorithm uses randomized leverage score sampling to approximately solve constrained least squares problems. Many successful methods for SymNMF rely on (approximately) solving sequences of constrained least squares problems. We prove theoretically that leverage score sampling can approximately solve nonnegative least squares problems to a chosen accuracy with high probability. Additionally, we prove sampling complexity results for previously proposed hybrid sampling techniques which deterministically include high leverage score rows. This hybrid scheme is crucial for obtaining speedups in practice. Finally, we demonstrate that both methods work well in practice by applying them to graph clustering tasks on large real world data sets. These experiments show that our methods approximately maintain solution quality and achieve significant speedups for both large dense and large sparse problems. 

   more » « less
5. On Sketching Approximations for Symmetric Boolean CSPs

   Boyland, Joanna; Hwang, Michael; Prasad, Tarun; Singer, Noah; Velusamy, Santhoshini (September 2022, Approximation, Randomization, and Combinatorial Optimization. Algorithms and Techniques (APPROX/RANDOM 2022))

   Chakrabarti, Amit; Swamy, Chaitanya (Ed.)

   A Boolean maximum constraint satisfaction problem, Max-CSP(f), is specified by a predicate f:{-1,1}^k → {0,1}. An n-variable instance of Max-CSP(f) consists of a list of constraints, each of which applies f to k distinct literals drawn from the n variables. For k = 2, Chou, Golovnev, and Velusamy [Chou et al., 2020] obtained explicit ratios characterizing the √ n-space streaming approximability of every predicate. For k ≥ 3, Chou, Golovnev, Sudan, and Velusamy [Chou et al., 2022] proved a general dichotomy theorem for √ n-space sketching algorithms: For every f, there exists α(f) ∈ (0,1] such that for every ε > 0, Max-CSP(f) is (α(f)-ε)-approximable by an O(log n)-space linear sketching algorithm, but (α(f)+ε)-approximation sketching algorithms require Ω(√n) space. In this work, we give closed-form expressions for the sketching approximation ratios of multiple families of symmetric Boolean functions. Letting α'_k = 2^{-(k-1)} (1-k^{-2})^{(k-1)/2}, we show that for odd k ≥ 3, α(kAND) = α'_k, and for even k ≥ 2, α(kAND) = 2α'_{k+1}. Thus, for every k, kAND can be (2-o(1))2^{-k}-approximated by O(log n)-space sketching algorithms; we contrast this with a lower bound of Chou, Golovnev, Sudan, Velingker, and Velusamy [Chou et al., 2022] implying that streaming (2+ε)2^{-k}-approximations require Ω(n) space! We also resolve the ratio for the "at-least-(k-1)-1’s" function for all even k; the "exactly-(k+1)/2-1’s" function for odd k ∈ {3,…,51}; and fifteen other functions. We stress here that for general f, the dichotomy theorem in [Chou et al., 2022] only implies that α(f) can be computed to arbitrary precision in PSPACE, and thus closed-form expressions need not have existed a priori. Our analyses involve identifying and exploiting structural "saddle-point" properties of this dichotomy. Separately, for all threshold functions, we give optimal "bias-based" approximation algorithms generalizing [Chou et al., 2020] while simplifying [Chou et al., 2022]. Finally, we investigate the √ n-space streaming lower bounds in [Chou et al., 2022], and show that they are incomplete for 3AND, i.e., they fail to rule out (α(3AND})-ε)-approximations in o(√ n) space. 

   more » « less