Massive Instances Research Articles

A Label Disambiguation-Based Multimodal Massive Multiple Instance Learning Approach for Immune Repertoire Classification

One individual human’s immune repertoire consists of a huge set of adaptive immune receptors at a certain time point, representing the individual's adaptive immune state. Immune repertoire classification and associated receptor identification have the potential to make a transformative contribution to the development of novel vaccines and therapies. The vast number of instances and exceedingly low witness rate pose a great challenge to the immune repertoire classification, which can be formulated as a Massive Multiple Instance Learning (MMIL) problem. Traditional MIL methods, at both bag-level and instance-level, confront the issues of substantial computational burden or supervision ambiguity when handling massive instances. To address these issues, we propose a novel label disambiguation-based multimodal massive multiple instance learning approach (LaDM³IL) for immune repertoire classification. LaDM³IL adapts the instance-level MIL paradigm to deal with the issue of high computational cost and employs a specially-designed label disambiguation module for label correction, mitigating the impact of misleading supervision. To achieve a more comprehensive representation of each receptor, LaDM³IL leverages a multimodal fusion module with gating-based attention and tensor-fusion to integrate the information from gene segments and amino acid (AA) sequences of each immune receptor. Extensive experiments on the Cytomegalovirus (CMV) and Cancer datasets demonstrate the superior performance of the proposed LaDM³IL for both immune repertoire classification and associated receptor identification tasks. The code is publicly available at https://github.com/Josie-xufan/LaDM3IL.

Fast Adaptive Non-Monotone Submodular Maximization Subject to a Knapsack Constraint

Constrained submodular maximization problems encompass a wide variety of applications, including personalized recommendation, team formation, and revenue maximization via viral marketing. The massive instances occurring in modern-day applications can render existing algorithms prohibitively slow. Moreover, frequently those instances are also inherently stochastic. Focusing on these challenges, we revisit the classic problem of maximizing a (possibly non-monotone) submodular function subject to a knapsack constraint. We present a simple randomized greedy algorithm that achieves a 5.83-approximation and runs in O(n log n) time, i.e., at least a factor n faster than other state-of-the-art algorithms. The versatility of our approach allows us to further transfer it to a stochastic version of the problem. There, we obtain a (9 + ε)-approximation to the best adaptive policy, which is the first constant approximation for non-monotone objectives. Experimental evaluation of our algorithms showcases their improved performance on real and synthetic data.

WITHDRAWN: Machine learning based implementation with apache storm for cavernous analytics of data

Big Data Analytics is one of the major study sectors with a variety of methods for data science and foresight. There are various situations in which big data is logged every day and a comprehensive research and development assessment is required. There are massive instances of care, review and forecasting of huge volumes of data being important routinely in the area of medical research. According to First Post reports data from each 500-bed hospital in the USA of more than 50 peta bytes are produced. Further analysis reveals that one gramme of DNA in digital medium amounts to 215 petabytes. The amount of smart wearable devices soared from 26 million in 2014 to more than 100 million in 2016, in another digital connectivity scenario.

Quadruply Stochastic Gradient Method for Large Scale Nonlinear Semi-Supervised Ordinal Regression AUC Optimization

Semi-supervised ordinal regression (S2OR) problems are ubiquitous in real-world applications, where only a few ordered instances are labeled and massive instances remain unlabeled. Recent researches have shown that directly optimizing concordance index or AUC can impose a better ranking on the data than optimizing the traditional error rate in ordinal regression (OR) problems. In this paper, we propose an unbiased objective function for S2OR AUC optimization based on ordinal binary decomposition approach. Besides, to handle the large-scale kernelized learning problems, we propose a scalable algorithm called QS3ORAO using the doubly stochastic gradients (DSG) framework for functional optimization. Theoretically, we prove that our method can converge to the optimal solution at the rate of O(1/t), where t is the number of iterations for stochastic data sampling. Extensive experimental results on various benchmark and real-world datasets also demonstrate that our method is efficient and effective while retaining similar generalization performance.

Large-scale graph generation: Recent results of the SPP 1736 – Part II

Abstract The selection of input data is a crucial step in virtually every empirical study. Experimental campaigns in algorithm engineering, experimental algorithmics, network analysis, and many other fields often require suited network data. In this context, synthetic graphs play an important role, as data sets of observed networks are typically scarce, biased, not sufficiently understood, and may pose logistic and legal challenges. Just like processing huge graphs becomes challenging in the big data setting, new algorithmic approaches are necessary to generate such massive instances efficiently. Here, we update our previous survey [35] on results for large-scale graph generation obtained within the DFG priority programme SPP 1736 (Algorithms for Big Data); to this end, we broaden the scope and include recently published results.

Distributed balanced color assignment on arbitrary networks

Consider a scenario in which a set of n agents hold items, where each item can be of one out of m possible types (colors). The agents are connected by an arbitrary network and have to distributively find a repartition of the colors such that the amount of colors for each agent is as balanced as possible: in the particular case where m is a multiple of n, each agent must have exactly m/n colors. More formally, the goal is to let the agents agree on an assignment of colors to agents such that the following two conditions hold:(i)each color is assigned to exactly one agent;(ii)each agent has at least ⌊m/n⌋ and at most ⌈m/n⌉ colors. Among all possible such repartitions, we seek for the one that minimizes the number of “changes” (measured in terms of misplaced items) with respect to the initial configuration. In other words, our aim is to design a distributed algorithm that finds a balanced coloring of minimum cost, where the cost of a coloring is the number of items that need to be relocated.This kind of questions, usually modeled as generalized bipartite matching problems, have been studied in the distributed setting only on clusters of commodity nodes with the aim of copying with real-world massive instances, which may involve millions of agents and colors. Here we propose the first distributed algorithm designed to run on an arbitrary network with the agents as nodes. Our algorithm turns out to be efficient in terms of both time and message complexity for a large class of graphs. Our results can be outlined as follows.We prove a lower bound Ω(m/n⋅D2) on message complexity, where D is the diameter of the graph, that holds for any approximation algorithm whose solution has cost at most 2(α−2)/α times the cost of any optimal solution, for every constant α>2. We give a distributed deterministic algorithm with time complexity O(max⁡{n2,Dlog⁡q}) and message complexity O(nlog⁡n⋅(log⁡q+mlog⁡m)), where q is the maximum number of items of a given color held by any agent. We show that the cost of our solution for arbitrary graphs is at most (2+δ) times the optimal cost, for any δ>0. We finally observe that, for large diameter graphs (i.e., D=Ω(nϵ), ϵ>0), we get matching lower and upper bounds on message complexity for the vast majority of instances of potential interest, that is, instances with polynomial number of colors and (up to) super-exponential number of items.

New heuristic approaches for maximum balanced biclique problem

The maximum balanced biclique problem (MBBP) is an important extension of the maximum clique problem (MCP), which has wide industrial applications. In this paper, we propose a new local search framework for MBBP where four heuristics are incorporated to improve its performance. Our framework alternates between an extension phase via adding vertex pairs and a restarting phase via removing vertex pairs. Three heuristics are proposed for selecting the pairs for addition and removal. The first heuristic is a prediction score function to greedily select the vertex pairs for addition, which makes use of the structural information of the problem. The second heuristic is a self-adaptive restarting heuristic that removes a dynamic number of vertex pairs from the candidate solution to allow the search to continue from a new search area. The third heuristic is proposed for solving massive graphs and is called the two-mode perturbation heuristic. It is used for selecting pairs of vertices for addition and lowers the average complexity for this task. We also introduce a k-bipartite core reduction rule to decrease the scale of all massive instances, which helps our algorithm find optimal solutions for many massive instances. These techniques lead to two efficient local search algorithms for MBBP. Experimental results demonstrate that the proposed algorithms can scale up to massive instances with billions of edges and that the proposed algorithms outperform state-of-the-art MBBP algorithms on standard benchmarks.

Scalable generation of scale-free graphs

We explain how massive instances of scale-free graphs following the Barabasi–Albert preferential attachment model can be generated very quickly in an embarrassingly parallel way. This makes this popular model available for studying big data graph problems. As a demonstration, we generated a Petaedge graph (1015 edges) in less than an hour.

A novel features ranking metric with application to scalable visual and bioinformatics data classification

Coming with the big data era, the filtering of uninformative data becomes emerging. To this end, ranking the high dimensionality features plays an important role. However, most of the state-of-art methods focus on improving the classification accuracy while the stability of the dimensionality reduction is simply ignored. In this paper, we proposed a Max-Relevance-Max-Distance (MRMD) feature ranking method, which balances accuracy and stability of feature ranking and prediction task. In order to prove the effectiveness on big data, we tested our method on two different datasets. The first one is image classification, which is a benchmark dataset with high dimensionality, while the second one is protein–protein interaction prediction data, which comes from our previous private research and has massive instances. Experiments prove that our method maintained the accuracy together with the stability on both two big datasets. Moreover, our method runs faster than other filtering and wrapping methods, such as mRMR and Information Gain.

A practical approximation algorithm for solving massive instances of hybridization number for binary and nonbinary trees.

BackgroundReticulate events play an important role in determining evolutionary relationships. The problem of computing the minimum number of such events to explain discordance between two phylogenetic trees is a hard computational problem. Even for binary trees, exact solvers struggle to solve instances with reticulation number larger than 40-50.ResultsHere we present CycleKiller and NonbinaryCycleKiller, the first methods to produce solutions verifiably close to optimality for instances with hundreds or even thousands of reticulations.ConclusionsUsing simulations, we demonstrate that these algorithms run quickly for large and difficult instances, producing solutions that are very close to optimality. As a spin-off from our simulations we also present TerminusEst, which is the fastest exact method currently available that can handle nonbinary trees: this is used to measure the accuracy of the NonbinaryCycleKiller algorithm. All three methods are based on extensions of previous theoretical work (SIDMA 26(4):1635-1656, TCBB 10(1):18-25, SIDMA 28(1):49-66) and are publicly available. We also apply our methods to real data.

Formation of gas discs in merging galaxies

Observations indicate that much of the interstellar gas in merging galaxies may settle into extended gaseous discs. Here, I present simulations of disc formation in mergers of gas-rich galaxies. Up to half of the total gas settles into embedded discs; the most massive instances result from encounters in which both galaxies are inclined to the orbital plane. These discs are often warped, many have rather complex kinematics, and roughly a quarter have counter-rotating or otherwise decoupled central components. Discs typically grow from the inside out; infall from tidal tails may continue disc formation over long periods of time.