Maximizing Submodular Function over the Integer LatticeTasuku Soma
The document describes generalizations of submodular function maximization and submodular cover problems from sets to integer lattices. It presents polynomial-time approximation algorithms for maximizing monotone diminishing return (DR) submodular functions subject to constraints like cardinality, polymatroid and knapsack on the integer lattice. It also presents an algorithm for the DR-submodular cover problem of minimizing cost subject to achieving a quality threshold. The results provide useful extensions of submodular optimization to settings that cannot be modeled as set functions.
The low-rank basis problem for a matrix subspaceTasuku Soma
This document summarizes a presentation on finding low-rank bases for matrix subspaces. It introduces the low-rank basis problem, describes a greedy algorithm to solve it using two phases - rank estimation and alternating projection, and proves local convergence guarantees for the algorithm. Experimental results on synthetic and image data demonstrate the algorithm can recover known low-rank bases and separate mixed images. Comparisons are made to tensor decomposition methods for the special case of rank-1 bases.
Fast Deterministic Algorithms for Matrix Completion ProblemsTasuku Soma
This document summarizes research on fast deterministic algorithms for matrix completion problems. It presents new algorithms for:
1) Matrix completion by rank-one matrices, solving it faster than previous work in O((m+n)2.77) time rather than O(m4.37n) time, where m is the larger matrix dimension and n is the number of indeterminates.
2) Mixed skew-symmetric matrix completion, the first deterministic polynomial time algorithm for this problem.
3) Skew-symmetric matrix completion by rank-two skew-symmetric matrices, the first deterministic polynomial time algorithm for this problem. The algorithms work over an arbitrary field.
Optimal Budget Allocation: Theoretical Guarantee and Efficient AlgorithmTasuku Soma
The document presents two main results:
1. A general framework for submodular function maximization over integer lattices with a (1-1/e)-approximation algorithm that runs in pseudo polynomial time. This extends budget allocation to more complex scenarios.
2. A faster algorithm for budget allocation when influence probabilities are non-increasing, running in almost linear time compared to previous polynomial time algorithms. Experiments on real and large synthetic graphs show it outperforms heuristics by up to 15%.
Regret Minimization in Multi-objective Submodular Function MaximizationTasuku Soma
This document presents algorithms for minimizing regret ratio in multi-objective submodular function maximization. It introduces the concept of regret ratio for evaluating the quality of a solution set for multiple objectives. It then proposes two algorithms, Coordinate and Polytope, that provide upper bounds on regret ratio by leveraging approximation algorithms for single objective problems. Experimental results on a movie recommendation dataset show the proposed algorithms achieve significantly lower regret ratios than a random baseline.
Maximizing Submodular Function over the Integer LatticeTasuku Soma
The document describes generalizations of submodular function maximization and submodular cover problems from sets to integer lattices. It presents polynomial-time approximation algorithms for maximizing monotone diminishing return (DR) submodular functions subject to constraints like cardinality, polymatroid and knapsack on the integer lattice. It also presents an algorithm for the DR-submodular cover problem of minimizing cost subject to achieving a quality threshold. The results provide useful extensions of submodular optimization to settings that cannot be modeled as set functions.
The low-rank basis problem for a matrix subspaceTasuku Soma
This document summarizes a presentation on finding low-rank bases for matrix subspaces. It introduces the low-rank basis problem, describes a greedy algorithm to solve it using two phases - rank estimation and alternating projection, and proves local convergence guarantees for the algorithm. Experimental results on synthetic and image data demonstrate the algorithm can recover known low-rank bases and separate mixed images. Comparisons are made to tensor decomposition methods for the special case of rank-1 bases.
Fast Deterministic Algorithms for Matrix Completion ProblemsTasuku Soma
This document summarizes research on fast deterministic algorithms for matrix completion problems. It presents new algorithms for:
1) Matrix completion by rank-one matrices, solving it faster than previous work in O((m+n)2.77) time rather than O(m4.37n) time, where m is the larger matrix dimension and n is the number of indeterminates.
2) Mixed skew-symmetric matrix completion, the first deterministic polynomial time algorithm for this problem.
3) Skew-symmetric matrix completion by rank-two skew-symmetric matrices, the first deterministic polynomial time algorithm for this problem. The algorithms work over an arbitrary field.
Optimal Budget Allocation: Theoretical Guarantee and Efficient AlgorithmTasuku Soma
The document presents two main results:
1. A general framework for submodular function maximization over integer lattices with a (1-1/e)-approximation algorithm that runs in pseudo polynomial time. This extends budget allocation to more complex scenarios.
2. A faster algorithm for budget allocation when influence probabilities are non-increasing, running in almost linear time compared to previous polynomial time algorithms. Experiments on real and large synthetic graphs show it outperforms heuristics by up to 15%.
Regret Minimization in Multi-objective Submodular Function MaximizationTasuku Soma
This document presents algorithms for minimizing regret ratio in multi-objective submodular function maximization. It introduces the concept of regret ratio for evaluating the quality of a solution set for multiple objectives. It then proposes two algorithms, Coordinate and Polytope, that provide upper bounds on regret ratio by leveraging approximation algorithms for single objective problems. Experimental results on a movie recommendation dataset show the proposed algorithms achieve significantly lower regret ratios than a random baseline.
Nonconvex Compressed Sensing with the Sum-of-Squares MethodTasuku Soma
This document presents a method for nonconvex compressed sensing using the sum-of-squares (SoS) method. It formulates q-minimization, which requires fewer samples than l1-minimization but is nonconvex, as a polynomial optimization problem. The SoS method is then applied to obtain a pseudoexpectation operator satisfying a pseudo robust null space property, guaranteeing stable signal recovery. Specifically, it shows that for a Rademacher measurement matrix, with the number of measurements scaling quadratically in the sparsity s, the SoS method finds a solution x^ satisfying ||x^-x||_q ≤ O(σs(x)q) + ε, providing nearly q-stable recovery.
Multicasting in Linear Deterministic Relay Network by Matrix CompletionTasuku Soma
This document presents a new algorithm for multicasting in linear deterministic relay networks (LDRNs) that is faster than previous algorithms. The algorithm works by first solving the unicast subproblems using an existing algorithm, then determining the linear encoding matrices for each layer simultaneously using mixed matrix completion. This allows the encoding matrices for an entire layer to be determined at once, rather than one node at a time. The new algorithm runs in O(dq(nr)^3 log(nr)) time, which is faster than the previous best algorithm when n = o(r).
StudySapuri Data Analytics Platform with Treasure DataTetsuo Yamabe
The document discusses Recruit Marketing Partners' migration of their StudySapuri online learning platform's data analytics capabilities to the Treasure Data platform. Key points include:
- StudySapuri previously used an on-premise Hadoop cluster for analytics but migrated to Treasure Data for a fully-managed cloud solution.
- The migration involved importing raw data from various sources into Treasure Data and transforming it using Luigi and Presto for reporting and insights.
- Challenges of the new platform include managing access across teams, handling large result files, and testing queries. Future work may involve machine learning and real-time data feeds.
The document is comprised of technical documentation copyrighted by Recruit Marketing Partners Co., Ltd. It includes code snippets and configuration examples for tools like AWS Lambda, Kinesis, Presto, Hive, Embulk and Treasure Data. The documentation provides guidance on building data pipelines, ETL processes, and reporting solutions using these technologies.
Nonconvex Compressed Sensing with the Sum-of-Squares MethodTasuku Soma
This document presents a method for nonconvex compressed sensing using the sum-of-squares (SoS) method. It formulates q-minimization, which requires fewer samples than l1-minimization but is nonconvex, as a polynomial optimization problem. The SoS method is then applied to obtain a pseudoexpectation operator satisfying a pseudo robust null space property, guaranteeing stable signal recovery. Specifically, it shows that for a Rademacher measurement matrix, with the number of measurements scaling quadratically in the sparsity s, the SoS method finds a solution x^ satisfying ||x^-x||_q ≤ O(σs(x)q) + ε, providing nearly q-stable recovery.
Multicasting in Linear Deterministic Relay Network by Matrix CompletionTasuku Soma
This document presents a new algorithm for multicasting in linear deterministic relay networks (LDRNs) that is faster than previous algorithms. The algorithm works by first solving the unicast subproblems using an existing algorithm, then determining the linear encoding matrices for each layer simultaneously using mixed matrix completion. This allows the encoding matrices for an entire layer to be determined at once, rather than one node at a time. The new algorithm runs in O(dq(nr)^3 log(nr)) time, which is faster than the previous best algorithm when n = o(r).
StudySapuri Data Analytics Platform with Treasure DataTetsuo Yamabe
The document discusses Recruit Marketing Partners' migration of their StudySapuri online learning platform's data analytics capabilities to the Treasure Data platform. Key points include:
- StudySapuri previously used an on-premise Hadoop cluster for analytics but migrated to Treasure Data for a fully-managed cloud solution.
- The migration involved importing raw data from various sources into Treasure Data and transforming it using Luigi and Presto for reporting and insights.
- Challenges of the new platform include managing access across teams, handling large result files, and testing queries. Future work may involve machine learning and real-time data feeds.
The document is comprised of technical documentation copyrighted by Recruit Marketing Partners Co., Ltd. It includes code snippets and configuration examples for tools like AWS Lambda, Kinesis, Presto, Hive, Embulk and Treasure Data. The documentation provides guidance on building data pipelines, ETL processes, and reporting solutions using these technologies.