Common Search Algorithms Research Articles

Conformational Sampling for Transition State Searches on a Computational Budget.

Transition state searches are the basis for computationally characterizing reaction mechanisms, making them a pivotal tool in myriad chemical applications. Nevertheless, common search algorithms are sensitive to reaction conformations, and the conformational spaces of even medium-sized reacting systems are too complex to explore with brute force. Here, we show that it is possible to train a classifier to learn the features of reaction conformers that conduce successful transition state searches, such that optimal conformers can be down-selected before incurring the cost of a high-level transition state search. The efficacy and transferability of this approach were tested using four distinct benchmarks comprising over three hundred individual reactions. Neglecting conformer contributions led to qualitatively incorrect activation energy estimations for a broad range of reactions, whereas simple random forest classifiers reliably down-selected low-barrier reaction conformers for unseen reactions. The robust performance of these machine learning classifiers mitigates cost as a factor when implementing conformational sampling into contemporary reaction prediction workflows and opens up many avenues for further improvements as transition state data grow.

A novel adaptive convolutional neural network for fault diagnosis of hydraulic piston pump with acoustic images

As an essential part of hydraulic transmission systems, hydraulic piston pumps have a significant role in many state-of-the-art industries. Thus, it is important to implement accurate and effective fault diagnosis of hydraulic piston pumps. Owing to the heavy reliance of shallow machine learning models on the expertise and experience of engineers, fault diagnosis based on deep models has attracted significant attention from academia and industry. To construct a deep model with good performance, it is necessary and challenging to tune the hyperparameters (HPs). Since many existing methods focus on manual tuning and use common search algorithms, it is meaningful to explore more intelligent algorithms that can automatically optimize the HPs. In this paper, Bayesian optimization (BO) is employed for adaptive HP learning, and an improved convolutional neural network (CNN) is established for fault feature extraction and classification in a hydraulic piston pump. First, acoustic signals are transformed into time–frequency distributions by a continuous wavelet transform. Second, a preliminary CNN model is built by setting initial HPs. The range of each HP to be optimized is identified. Third, BO is employed to select the optimal combination of HPs. An improved model called CNN-BO is constructed. Finally, the diagnostic efficiency of CNN-BO is analyzed using a confusion matrix and t-distributed stochastic neighbor embedding. The classification performance of different models is compared. It is found that CNN-BO has a higher accuracy and better robustness in fault diagnosis for a hydraulic piston pump. This research will provide a basis for ensuring the reliability and safety of the hydraulic pump.

Effective rural electrification via optimal network: Optimal path-finding in highly anisotropic search space using multiplier-accelerated A* algorithm

United Nations’ 7th Sustainable Development Goal envisions the availability of modern energy for everyone by 2030. While the progress has been satisfactory in the last few years, further rural electrification is increasingly challenging. The current mainstream approach of electrifying villages individually is becoming cost-ineffective due to uncertainties in both resource availability and energy demand for small, difficult-to-reach, residences. A networked rural electrification model, i.e. a cost-optimized network connecting villages and generation facilities, could improve resources utilization, reliability and flexibility. However, determining optimal paths with common search algorithms is extremely inefficient due to complex topographic features of rural areas. This work develops and applies an artificial intelligence search method to efficiently route inter-village power connections in the common rural electrification situation where substantial topological variations exist. The method is evolved from the canonical A* algorithm. Results compare favorably with optimal A* results, at significantly reduced computational effort. Furthermore, users can adaptively trade-off between computation speed and optimality and hence quickly evaluate sites and configurations at reasonable accuracy, which is impossible with classical methods.

Sleep Sets Meet Duplicate Elimination

The sleep sets technique is a path-dependent pruning method for state space search. In the past, the combination of sleep sets with graph search algorithms that perform duplicate elimination has often shown to be error-prone. In this paper, we provide the theoretical basis for the integration of sleep sets with common search algorithms in AI that perform duplicate elimination. Specifically, we investigate approaches to safely integrate sleep sets with optimal (best-first) search algorithms. Based on this theory, we provide an initial step towards integrating sleep sets within A* and additional state pruning techniques like strong stubborn sets. Our experiments show slight, yet consistent improvements on the number of generated search nodes across a large number of standard domains from the international planning competitions.

Constrained Symbolic Search: On Mutexes, BDD Minimization and More

Symbolic search allows saving large amounts of memory compared to regular explicit-state search algorithms. This is crucial in optimal settings, in which common search algorithms often exhaust the available memory. So far, the most successful uses of symbolic search have been bidirectional blind search and the generation of abstraction heuristics like Pattern Databases. Despite its usefulness, several common techniques in explicit-state search have not been employed in symbolic search. In particular, mutexes and other constraining invariants, techniques that have been proven essential when doing regression, are yet to be exploited in conjunction with BDDs. In this paper we analyze the use of such constraints in symbolic search and its combination with minimization techniques common in BDD manipulation. Experimental results show a significant increase in performance, considerably above the current state of the art in optimal planning.

Microbial inositol polyphosphate metabolic pathway as drug development target.

Inositol polyphosphates are a diverse and multifaceted class of intracellular messengers omnipresent in eukaryotic cells. These water-soluble molecules regulate many aspects of fundamental cell physiology. Removing this metabolic pathway is deleterious: inositol phosphate kinase null mutations can result in lethality or substantial growth phenotypes. Inositol polyphosphate synthesis occurs through the actions of a set of kinases that phosphorylate phospholipase-generated IP3 to higher phosphorylated forms, such as the fully phosphorylated IP6 and the inositol pyrophosphates IP7 and IP8. Unicellular organisms have a reduced array of the kinases for synthesis of higher phosphorylated inositol polyphosphates, while human cells possess two metabolic routes to IP6. The enzymes responsible for inositol polyphosphate synthesis have been identified in all eukaryote genomes, although their amino acid sequence homology is often barely detectable by common search algorithms. Homology between human and microbial inositol phosphate kinases is restricted to a few catalytically important residues. Recent studies of the inositol phosphate metabolic pathways in pathogenic fungi (Cryptococcus neoformans) and protozoa (Trypanosome brucei) have revealed the importance of the highly phosphorylated inositol polyphosphates to the fitness and thus virulence of these pathogens. Given this, identification of inositol kinase inhibitors specifically targeting the kinases of pathogenic microorganisms is desirable and achievable.

PhosSA: Fast and accurate phosphorylation site assignment algorithm for mass spectrometry data

Phosphorylation site assignment of high throughput tandem mass spectrometry (LC-MS/MS) data is one of the most common and critical aspects of phosphoproteomics. Correctly assigning phosphorylated residues helps us understand their biological significance. The design of common search algorithms (such as Sequest, Mascot etc.) do not incorporate site assignment; therefore additional algorithms are essential to assign phosphorylation sites for mass spectrometry data. The main contribution of this study is the design and implementation of a linear time and space dynamic programming strategy for phosphorylation site assignment referred to as PhosSA. The proposed algorithm uses summation of peak intensities associated with theoretical spectra as an objective function. Quality control of the assigned sites is achieved using a post-processing redundancy criteria that indicates the signal-to-noise ratio properties of the fragmented spectra. The quality assessment of the algorithm was determined using experimentally generated data sets using synthetic peptides for which phosphorylation sites were known. We report that PhosSA was able to achieve a high degree of accuracy and sensitivity with all the experimentally generated mass spectrometry data sets. The implemented algorithm is shown to be extremely fast and scalable with increasing number of spectra (we report up to 0.5 million spectra/hour on a moderate workstation). The algorithm is designed to accept results from both Sequest and Mascot search engines. An executable is freely available at http://helixweb.nih.gov/ESBL/PhosSA/ for academic research purposes.

A SNOMED supported ontological vector model for subclinical disorder detection using EHR similarity

Electronic Health Records (EHR) form a valuable resource in the healthcare enterprise because clinical evidence can be provided to identify potential complications and support decisions on early intervention. Simple string matching, the common search algorithm, is not able to map a query to the similar health records in the database with respect to the medical concepts. A novel ontological vector model supported by the Systematized Nomenclature of Medicine Clinical Terms (SNOMED-CT) is proposed in this paper to project the disease terms of a health record to a feature space so that each health record can be characterized using a feature vector, giving a fingerprint of the record. The similarity between the query and database health records was measured by similarity measures of their feature vectors and string matching score respectively. Three types of similarity measures were considered in this study, namely, Euclidean distance (ED), direction cosine (DC) and modified direction cosine (mDC). Medical history and carotid ultrasonic imaging findings were collected from 47 subjects in Hong Kong. The dataset formed 1081 pairs of health records and ROC analysis was used to evaluate and compare the accuracy of the ontological vector model and simple string matching against the agreement of the presence or absence of carotid plaques identified by carotid ultrasound between two subjects. It was found that the score generated by simple string matching was a random rater but the ontological vector model was not. In other words, the degree of health record similarity based on the ontological vector model is associated with the agreement of atherosclerosis between two patients. The vector model using feature terms at the SNOMED-CT level 4 gave the best performance. The performance of mDC was very close to that of ED and DC but the properties of mDC make it more suitable for the retrieval of similar health records. It was also shown that the ontological vector model was enhanced by the support vector classifier approach.

Applying Soft Arc Consistency to Distributed Constraint Optimization Problems

The Distributed Constraint Optimization Problem (DCOP) is a fundamental framework of multi-agent systems. With DCOPs a multi-agent system is represented as a set of variables and a set of constraints/cost functions. Distributed task scheduling and distributed resource allocation can be formalized as DCOPs. In this paper, we propose an efficient method that applies directed soft arc consistency to a DCOP. In particular, we focus on DCOP solvers that employ pseudo-trees. A pseudo-tree is a graph structure for a constraint network that represents a partial ordering of variables. Some pseudo-tree-based search algorithms perform optimistic searches using explicit/implicit backtracking in parallel. However, for cost functions taking a wide range of cost values, such exact algorithms require many search iterations. Therefore additional improvements are necessary to reduce the number of search iterations. A previous study used a dynamic programming-based preprocessing technique that estimates the lower bound values of costs. However, there are opportunities for further improvements of efficiency. In addition, modifications of the search algorithm are necessary to use the estimated lower bounds. The proposed method applies soft arc consistency (soft AC) enforcement to DCOP. In the proposed method, directed soft AC is performed based on a pseudo-tree in a bottom up manner. Using the directed soft AC, the global lower bound value of cost functions is passed up to the root node of the pseudo-tree. It also totally reduces values of binary cost functions. As a result, the original problem is converted to an equivalent problem. The equivalent problem is efficiently solved using common search algorithms. Therefore, no major modifications are necessary in search algorithms. The performance of the proposed method is evaluated by experimentation. The results show that it is more efficient than previous methods.

Detection of illicit substances in fingerprints by infrared spectral imaging

FTIR and Raman spectral imaging can be used to simultaneously image a latent fingerprint and detect exogenous substances deposited within it. These substances might include drugs of abuse or traces of explosives or gunshot residue. In this work, spectral searching algorithms were tested for their efficacy in finding targeted substances deposited within fingerprints. "Reverse" library searching, where a large number of possibly poor-quality spectra from a spectral image are searched against a small number of high-quality reference spectra, poses problems for common search algorithms as they are usually implemented. Out of a range of algorithms which included conventional Euclidean distance searching, the spectral angle mapper (SAM) and correlation algorithms gave the best results when used with second-derivative image and reference spectra. All methods tested gave poorer performances with first derivative and undifferentiated spectra. In a search against a caffeine reference, the SAM and correlation methods were able to correctly rank a set of 40 confirmed but poor-quality caffeine spectra at the top of a dataset which also contained 4,096 spectra from an image of an uncontaminated latent fingerprint. These methods also successfully and individually detected aspirin, diazepam and caffeine that had been deposited together in another fingerprint, and they did not indicate any of these substances as a match in a search for another substance which was known not to be present. The SAM was used to successfully locate explosive components in fingerprints deposited on silicon windows. The potential of other spectral searching algorithms used in the field of remote sensing is considered, and the applicability of the methods tested in this work to other modes of spectral imaging is discussed.

Dynamic Control in Real-Time Heuristic Search

Real-time heuristic search is a challenging type of agent-centered search because the agent's planning time per action is bounded by a constant independent of problem size. A common problem that imposes such restrictions is pathfinding in modern computer games where a large number of units must plan their paths simultaneously over large maps. Common search algorithms (e.g., A*, IDA*, D*, ARA*, AD*) are inherently not real-time and may lose completeness when a constant bound is imposed on per-action planning time. Real-time search algorithms retain completeness but frequently produce unacceptably suboptimal solutions. In this paper, we extend classic and modern real-time search algorithms with an automated mechanism for dynamic depth and subgoal selection. The new algorithms remain real-time and complete. On large computer game maps, they find paths within 7% of optimal while on average expanding roughly a single state per action. This is nearly a three-fold improvement in suboptimality over the existing state-of-the-art algorithms and, at the same time, a 15-fold improvement in the amount of planning per action.