AI Domain Research Articles

IntrospectiveQ-learning and learning from demonstration

AbstractOne challenge faced by reinforcement learning (RL) agents is that in many environments the reward signal is sparse, leading to slow improvement of the agent’s performance in early learning episodes. Potential-based reward shaping can help to resolve the aforementioned issue of sparse reward by incorporating an expert’s domain knowledge into the learning through a potential function. Past work on reinforcement learning from demonstration (RLfD) directly mapped (sub-optimal) human expert demonstration to a potential function, which can speed up RL. In this paper we propose an introspective RL agent that significantly further speeds up the learning. An introspective RL agent records its state–action decisions and experience during learning in a priority queue. Good quality decisions, according to a Monte Carlo estimation, will be kept in the queue, while poorer decisions will be rejected. The queue is then used as demonstration to speed up RL via reward shaping. A human expert’s demonstration can be used to initialize the priority queue before the learning process starts. Experimental validation in the 4-dimensional CartPole domain and the 27-dimensional Super Mario AI domain shows that our approach significantly outperforms non-introspective RL and state-of-the-art approaches in RLfD in both domains.

FB-STEP: A fuzzy Bayesian network based data-driven framework for spatio-temporal prediction of climatological time series data

With the recent development of computational intelligence (CI), data-driven models have gained growing interest to be applied in various scientific disciplines. This paper aims at proposing a hybrid CI-based data-driven framework as a complement for the physics-based models used in climatological prediction. The proposed framework, called FB-STEP, is based on a combination of fuzzy Bayesian strategy and multifractal analysis technique. The focus is to address three major research challenges in multivariate climatological prediction: (1) modeling complex spatio-temporal dependency among climatological variables, (2) dealing with non-linear, chaotic dynamics in climatic time series, and (3) reducing epistemic uncertainty in the data-driven prediction process. The present work not only explores Fuzzy-Bayesian modeling of spatio-temporal processes, but also presents an elegant approach of dealing with intrinsic chaos in time series, through a synergism between multifractal analysis and Bayesian inference mechanism. Similar concepts may also be successfully employed in developing expert or intelligent systems for wide range of applications, including reservoir-water dynamics modeling, flood monitoring, traffic flow modeling, chemical-mechanical process monitoring, and so on. Thus, the present research work carries a significant value not merely in the field of climate research, but also in the domains of AI and machine intelligence. The experimentation has been carried out to spatio-temporally extrapolate the climatic conditions of five different locations in India, with the help of historical data on temperature, humidity, precipitation rate, and soil moisture. A comparative study with popular linear and non-linear methods has validated the efficacy of the proposed data-driven approach for climatological prediction.

Swapping the N- and C-terminal domains of human apolipoprotein E3 and AI reveals insights into their structure/activity relationship.

Apolipoprotein (apo) E3 and apoAI are exchangeable apolipoproteins that play a dominant role in regulating plasma lipoprotein metabolism. ApoE3 (299 residues) is composed of an N-terminal (NT) domain bearing a 4-helix bundle and a C-terminal (CT) domain bearing a series of amphipathic α-helices. ApoAI (243 residues) also comprises a highly helical NT domain and a less structured CT tail. The objective of this study was to understand their structural and functional role by generating domain swapped chimeras: apoE3-NT/apoAI-CT and apoAI-NT/apoE-CT. The bacterially overexpressed chimeras were purified by affinity chromatography and their identity confirmed by immunoblotting and mass spectrometry. Their α-helical content was comparable to that of the parent proteins. ApoE3-NT/apoAI-CT retained the denaturation profile of apoE3 NT domain, with apoAI CT tail eliciting a relatively unstructured state; its lipid binding ability improved dramatically compared to apoE3 indicative of a significant role of apoAI CT tail in lipid binding interaction. The LDL receptor interaction and ability to promote ABCA1-mediated cholesterol efflux of apoE3-NT/apoAI-CT was comparable to that of apoE3. In contrast, apoAI-NT/apoE-CT elicited an unfolding pattern and lipid binding ability that were similar to that of apoAI. As expected, DMPC/apoAI-NT/apoE-CT discoidal particles did not elicit LDLr binding ability, and promoted SR-B1 mediated cellular uptake of lipids to a limited extent. However, apoAI-NT/apoE-CT displayed an enhanced ability to promote cholesterol efflux compared to apoAI, indicative of a significant role for apoE CT domain in mediating this function. Together, these results indicate that the functional attributes of apoAI and apoE3 can be conferred on each other and that NT-CT domain interactions significantly modulate their structure and function.

Artificial intelligence meeting neuropsychology. Semantic memory in normal and pathological aging

Artificial intelligence (IA) is the subject of much research, but also many fantasies. It aims to reproduce human intelligence in its learning capacity, knowledge storage and computation. In 2014, the Defense Advanced Research Projects Agency (DARPA) started the restoring active memory (RAM) program that attempt to develop implantable technology to bridge gaps in the injured brain and restore normal memory function to people with memory loss caused by injury or disease. In another IA's field, computational ontologies (a formal and shared conceptualization) try to model knowledge in order to represent a structured and unambiguous meaning of the concepts of a target domain. The aim of these structures is to ensure a consensual understanding of their meaning and a univariant use (the same concept is used by all to categorize the same individuals). The first representations of knowledge in the AI's domain are largely based on model tests of semantic memory. This one, as a component of long-term memory is the memory of words, ideas, concepts. It is the only declarative memory system that resists so remarkably to the effects of age. In contrast, non-specific cognitive changes may decrease the performance of elderly in various events and instead report difficulties of access to semantic representations that affect the semantics stock itself. Some dementias, like semantic dementia and Alzheimer's disease, are linked to alteration of semantic memory. We propose in this paper, using the computational ontologies model, a formal and relatively thin modeling, in the service of neuropsychology: 1) for the practitioner with decision support systems, 2) for the patient as cognitive prosthesis outsourced, and 3) for the researcher to study semantic memory.

A Novel Line Segments Extraction Algorithm with Variant Constraints

In the domain of CAM and AI, line segment extraction algorithm play an important role. In order to extract line segments in different degree of curvature and continuum, construct the flexible and variant linearization and continuum constraints by the array signal processing method. And through the setting dimension of image unit, extract line segments satisfying the constraints in different degrees. At the last some experiments on true images demonstrate the roles of flexible parameters and the efficiency of the algorithm.

Cysteine-rich domain of scavenger receptor AI modulates the efficacy of surface targeting and mediates oligomeric Aβ internalization

BackgroundInsufficient clearance of soluble oligomeric amyloid-β peptide (oAβ) in the central nervous system leads to the synaptic and memory deficits in Alzheimer's disease (AD). Previously we have identified scavenger receptor class A (SR-A) of microglia mediates oligomeric amyloid-β peptide (oAβ) internalization by siRNA approach. SR-A is a member of cysteine-rich domain (SRCR) superfamily which contains proteins actively modulating the innate immunity and host defense, however the functions of the SRCR domain remain unclear. Whether the SRCR domain of SR-AI modulates the receptor surface targeting and ligand internalization was investigated by expressing truncated SR-A variants in COS-7 cells. Surface targeting of SR-A variants was examined by live immunostaining and surface biotinylation assays. Transfected COS-7 cells were incubated with fluorescent oAβ and acetylated LDL (AcLDL) to assess their ligand-internalization capabilities.ResultGenetic ablation of SR-A attenuated the internalization of oAβ and AcLDL by microglia. Half of oAβ-containing endocytic vesicles was SR-A positive in both microglia and macrophages. Clathrin and dynamin in SR-AI-mediated oAβ internalization were involved. The SRCR domain of SR-AI is encoded by exons 10 and 11. SR-A variants with truncated exon 11 were intracellularly retained, whereas SR-A variants with further truncations into exon 10 were surface-targeted. The fusion of exon 11 to the surface-targeted SR-A variant lacking the SRCR domain resulted in the intracellular retention and the co-immunoprecipitation of Bip chaperon of the endoplasmic reticulum. Surface-targeted variants were N-glycosylated, whereas intracellularly-retained variants retained in high-mannose states. In addition to the collagenous domain, the SRCR domain is a functional binding domain for oAβ and AcLDL. Our data suggest that inefficient folding of SR-AI variants with truncated SRCR domain was recognized by the endoplasmic reticulum associated degradation which leads to the immature N- glycosylation and intracellular retention.ConclusionThe novel functions of the SRCR domain on regulating the efficacy of receptor trafficking and ligand binding may lead to possible approaches on modulating the innate immunity in Alzheimer’s disease and atherosclerosis.

Feature learning and deep architectures: new directions for music informatics

As we look to advance the state of the art in content-based music informatics, there is a general sense that progress is decelerating throughout the field. On closer inspection, performance trajectories across several applications reveal that this is indeed the case, raising some difficult questions for the discipline: why are we slowing down, and what can we do about it? Here, we strive to address both of these concerns. First, we critically review the standard approach to music signal analysis and identify three specific deficiencies to current methods: hand-crafted feature design is sub-optimal and unsustainable, the power of shallow architectures is fundamentally limited, and short-time analysis cannot encode musically meaningful structure. Acknowledging breakthroughs in other perceptual AI domains, we offer that deep learning holds the potential to overcome each of these obstacles. Through conceptual arguments for feature learning and deeper processing architectures, we demonstrate how deep processing models are more powerful extensions of current methods, and why now is the time for this paradigm shift. Finally, we conclude with a discussion of current challenges and the potential impact to further motivate an exploration of this promising research area.

Research on Product Modeling Method for Supporting the Networked Manufacturing

This paper introduces a new product modeling method: MAPMNM(Multi-Agent Based Product Modeling for Networked Manufacturing) based on analysing the requirements of product modeling. The method offers a unified product master model and distributed domain models to alliance enterprise based on network by using web service and the agent technology of AI domain. Firstly, the paper gives the framework of the MAPMNM.Secondly, it elaborates on the web services, the architecture and running menchanism of the domain agent in MAPMNM.

Beta-radiation reduces the reactivity of extra cellular matrix proteins in intravascular brachytherapy (IVBT), resulting in decreased platelet adhesion

BackgroundIntravascular Brachytherapy as a tool to reduce restenosis is thought to alter vascular wall biology and vessel wall protein function. Platelet accumulation is also indeed important in the genesis of restenosis.We examine the in vitro effects of beta-radiation on the certain vessel wall extra cellular matrix proteins. We hypothesized that vessel wall (proteins) had become less prone to thrombosis. MethodsWe examined platelet adhesion to 20-Gy beta radiation treated extra cellular matrix proteins under flow conditions. Platelet flow adhesion was evaluated or quantified by image analysis, aggregation size analysis using the Watershed program and real-time fluorescence images of thrombus formation. The effect of beta radiation on vWF was further showing by measuring the binding of domain-specific antibodies to radiation treated vWF. Results20-Gy beta radiation significantly decreased platelet adhesion to extra cellular matrix protein; vWF and collagen Type III and had no effect on the adhesion upon fibrinogen and fibronectin. The beta-radiation affected mostly the AI, A2 and A3 domains of the vWF molecule on the surface, whereas the D′–D3 and B–C1 domains on the surface remain unaffected and suggesting a significant decrease in vWF binding capacity to the GPIb, heparin and collagen ligands. ConclusionBeta radiation treatment can alter the reactivity of the certain vessel wall extra cellular matrix proteins, in particular vWF and collagen. The vessel wall may become less prone to platelet adhesion, which results in decrease thrombus formation. It might help to reduce the onset of acute coronary occlusion after the intervention.

Anti-infective activity of apolipoprotein domain derived peptides in vitro: identification of novel antimicrobial peptides related to apolipoprotein B with anti-HIV activity

BackgroundPrevious reports have shown that peptides derived from the apolipoprotein E receptor binding region and the amphipathic α-helical domains of apolipoprotein AI have broad anti-infective activity and antiviral activity respectively. Lipoproteins and viruses share a similar cell biological niche, being of overlapping size and displaying similar interactions with mammalian cells and receptors, which may have led to other antiviral sequences arising within apolipoproteins, in addition to those previously reported. We therefore designed a series of peptides based around either apolipoprotein receptor binding regions, or amphipathic α-helical domains, and tested these for antiviral and antibacterial activity.ResultsOf the nineteen new peptides tested, seven showed some anti-infective activity, with two of these being derived from two apolipoproteins not previously used to derive anti-infective sequences. Apolipoprotein J (151-170) - based on a predicted amphipathic alpha-helical domain from apolipoprotein J - had measurable anti-HSV1 activity, as did apolipoprotein B (3359-3367) dp (apoBdp), the latter being derived from the LDL receptor binding domain B of apolipoprotein B. The more active peptide - apoBdp - showed similarity to the previously reported apoE derived anti-infective peptide, and further modification of the apoBdp sequence to align the charge distribution more closely to that of apoEdp or to introduce aromatic residues resulted in increased breadth and potency of activity. The most active peptide of this type showed similar potent anti-HIV activity, comparable to that we previously reported for the apoE derived peptide apoEdpL-W.ConclusionsThese data suggest that further antimicrobial peptides may be obtained using human apolipoprotein sequences, selecting regions with either amphipathic α-helical structure, or those linked to receptor-binding regions. The finding that an amphipathic α-helical region of apolipoprotein J has antiviral activity comparable with that for the previously reported apolipoprotein AI derived peptide 18A, suggests that full-length apolipoprotein J may also have such activity, as has been reported for full-length apolipoprotein AI. Although the strength of the anti-infective activity of the sequences identified was limited, this could be increased substantially by developing related mutant peptides. Indeed the apolipoprotein B-derived peptide mutants uncovered by the present study may have utility as HIV therapeutics or microbicides.

Interactive Cost Configuration Over Decision Diagrams

In many AI domains such as product configuration, a user should interactively specify a solution that must satisfy a set of constraints. In such scenarios, offline compilation of feasible solutions into a tractable representation is an important approach to delivering efficient backtrack-free user interaction online. In particular,binary decision diagrams (BDDs) have been successfully used as a compilation target for product and service configuration. In this paper we discuss how to extend BDD-based configuration to scenarios involving cost functions which express user preferences. We first show that an efficient, robust and easy to implement extension is possible if the cost function is additive, and feasible solutions are represented using multi-valued decision diagrams (MDDs). We also discuss the effect on MDD size if the cost function is non-additive or if it is encoded explicitly into MDD. We then discuss interactive configuration in the presence of multiple cost functions. We prove that even in its simplest form, multiple-cost configuration is NP-hard in the input MDD. However, for solving two-cost configuration we develop a pseudo-polynomial scheme and a fully polynomial approximation scheme. The applicability of our approach is demonstrated through experiments over real-world configuration models and product-catalogue datasets. Response times are generally within a fraction of a second even for very large instances.

A PI3 Kinase Inhibitor Found to Activate Bestrophin 3

Bestrophin 3 (Best3), a member of bestrophin Cl channel family, is a CaCl(cGMP) channel candidate in vascular smooth muscle cells. The mechanism for its activation remains unclear. In previous studies, we reported that an autoinhibitory domain ((356)IPSFLGS(362)) existed in Best3 C-terminus and when the autoinhibitory domain was mutated, the Best3 channel was dramatically activated. In this study, we further dissected the roles of the C-terminal sequence in Best3 activation. We found that there were eight basic amino acids downstream of the AI domain within the region (384-397), which were also involved in Best3 activation. Mutations of these basic amino acids significantly activated Best3 as a Cl channel. Led by the assumption that the basic amino acids may be involved in the Best3 C-terminal membrane association through binding to membranous phospholipids, we discovered that PI3Kalpha inhibitor IV could strongly activate Best3.

Weighted [formula omitted] search – unifying view and application

The A ∗ algorithm is a well-known heuristic best-first search method. Several performance-accelerated extensions of the exact A ∗ approach are known. Interesting examples are approximate algorithms where the heuristic function used is inflated by a weight (often referred to as weighted A ∗ ). These methods guarantee a bounded suboptimality. As a technical contribution, this paper presents the previous results related to weighted A ∗ from authors like Pohl, Pearl, Kim, Likhachev and others in a more condensed and unifying form. With this unified view, a novel general bound on suboptimality of the result is derived. In the case of avoiding any reopening of expanded states, for ϵ > 0 , this bound is ( 1 + ϵ ) ⌊ N 2 ⌋ where N is an upper bound on an optimal solution length. Binary Decision Diagrams (BDDs) are well-known to AI, e.g. from set-based exploration of sparse-memory and symbolic manipulation of state spaces. The problem of exact or approximate BDD minimization is introduced as a possible new challenge for heuristic search. Like many classical AI domains, this problem is motivated by real-world applications. Several variants of weighted A ∗ search are applied to problems of BDD minimization and the more classical domains like blocksworld and sliding-tile puzzles. For BDD minimization, the comparison of the evaluated methods also includes previous heuristic and simulation-based methods such as Rudell's hill-climbing based sifting algorithm, Simulated Annealing and Evolutionary Algorithms. A discussion of the results obtained in the different problem domains gives our experiences with weighted A ∗ , which is of value for the AI practitioner.

A new chemiluminescence paradox: selective inhibition of isoluminol‐amplified activity in phagocytes by peptides from annexin AI

Chemiluminescence systems enhanced by either isoluminol or luminol in combination with a peroxidase are sensitive methods for the detection of reactive oxygen species (ROS) generated by phagocyte NADPH oxidase. The two amplifying substrates are structurally very similar, differing only in the position of the amino group in the aromatic ring of the molecules. This difference renders isoluminol a less lipophilic molecule that is less permeable to biological membranes. The use of isoluminol is consequently restricted to studies dealing with the secretion of oxygen metabolites. In this study we show that synthetic peptides derived from the N-terminal domain of the calcium-regulated protein annexin AI interfere with the detection of radicals in an isoluminol-amplified, but not in a luminol-amplified, system. The annexin AI-derived peptides reduce the light output with isoluminol excited by superoxide and horseradish peroxidase (HRP) in formyl-methionyl-leucyl-phenylalanine- and phorbol myristate acetate-stimulated cells, as well as by hydrogen peroxide and HRP. The precise mechanism for the inhibition is not known. The results presented strongly suggest that a reduced cellular response detected with isoluminol-amplified chemiluminescence should be confirmed with an alternative technique to determine release of superoxide anions and hydrogen peroxide.

Embedding Computational Intelligence in Pervasive Spaces

This paper presents how to embed AI mechanisms in pervasive spaces to produce more intelligent, adaptive, and convenient environments. The paper also concentrated on computational intelligence in particular, fuzzy systems and neural networks because this emerging AI domain covers most of the requirements for this task.

AI Meets Web 2.0: Building the Web of Tomorrow, Today

Imagine an Internet-scale knowledge system where people and intelligent agents can collaborate on solving complex problems in business, engineering, science, medicine, and other endeavors. Its resources include semantically tagged websites, wikis, and blogs, as well as social networks, vertical search engines, and a vast array of web services from business processes to AI planners and domain models. Research prototypes of decentralized knowledge systems have been demonstrated for years, but now, thanks to the web and Moore's law, they appear ready for prime time. This article introduces the architectural concepts for incrementally growing an Internet-scale knowledge system and illustrates them with scenarios drawn from e-commerce, e-science, and e-life.

Concurrent Java Test Generation as a Search Problem

A Random test generator generates executable tests together with their expected results. In the form of a noise-maker, it seeds the program with conditional scheduling primitives (such as yield()) that may cause context switches. As a result different interleavings are potentially produced in different executions of the program. Determining a-priori the set of seeded locations required for a bug to manifest itself is rarely possible.This work proposes to reformulate random test generation of concurrent Java programs as a search problem. Hence, it allows applying a set of well known search techniques from the domain of AI to the input space of the test generator. By iteratively refining the input parameters fed to the test generator, the search process creates testing scenarios (i.e. interleavings) that maximizes predefined objective functions. We develop geneticFinder, a noise-maker that uses a genetic algorithm as a search method. We demonstrate our approach by maximizing two objective functions: the high manifestation rate of concurrent bugs and the exporting of a high degree of debugging information to the user. Experimental results show our approach is effective.

Cortical convergence from different frequency domains in the cat primary auditory cortex

Primary auditory cortex (AI) has a tonotopic map consisting of orderly isofrequency (IF) bands, and cortical connections are commonly supposed to link domains preferring similar characteristic frequencies (CFs) within AI and in auditory association cortex. The interaction of different frequency channels, however, has not fully been understood in terms of anatomical substrates. Here, by injecting two anterograde tracers in different frequency domains of cat AI, without overlap of the injection cores, we attempted to relate the anatomical mapping of cortical outputs to physiologically defined fields in the auditory cortex. Consistent with previous studies, patches of labeled axon terminals were oriented largely along the IF axis. In regions distant from the injection sites, however, terminal patches were divergent in distribution. This divergence resulted in a complex geometry of partial overlap of projections originating from the two injection sites. The relative extent of the overlap tended to vary depending on the distance between the two injection sites. Physiological mapping for tonotopy across auditory fields revealed that projectional overlap was characteristic of dorsal AI and the dorsoposterior field and, to a lesser extent, in the secondary auditory field. Considering the differences in frequency representation in different AI IF bands, the anatomical convergence of projections tuned to different CFs could contribute to the spectral integration of sound components. Furthermore, the different extent of convergence in the functionally distinct fields might reflect field-specific processing of acoustic signals.

Artificial life and the Chinese room argument.

"Strong artificial life" refers to the thesis that a sufficiently sophisticated computer simulation of a life form is a life form in its own right. Can John Searle's Chinese room argument [12]-originally intended by him to show that the thesis he dubs "strong AI" is false-be deployed against strong ALife? We have often encountered the suggestion that it can be (even in print; see Harnad [8]). We do our best to transfer the argument from the domain of AI to that of ALife. We do so in order to show once and for all that the Chinese room argument proves nothing about ALife. There may indeed be powerful philosophical objections to the thesis of strong ALife, but the Chinese room argument is not among them.

ApoA-II Versus ApoA-I

Apolipoprotein A-I (apoA-I), the major protein of HDL, is probably among the most intensively studied of all proteins. It functions in HDL assembly, in the removal of excess cholesterol from cells, and in the transport of cholesterol from peripheral tissues to the liver, a process known as “reverse cholesterol transport.” It is a cofactor for lecithin:cholesterol acyl transferase (LCAT), the enzyme responsible for cholesterol esterification in HDL, and it stabilizes certain antioxidant enzymes, such as serum paraoxonase, that are carried on HDL.1,2 Rare null mutations of apoA-I in human populations are associated with early coronary heart disease,3 and overexpression of apoA-I in transgenic mice protects against atherogenesis.4 An exciting recent development was the 4-angstrom-resolution x-ray structure for the helical part of apoA-I, encompassing residues 44 to 243 (reviewed by Segrest et al5). See page 1977 In contrast, apolipoprotein A-II (apoA-II), the second most abundant protein of HDL, has no known function, and it has been the subject of few detailed studies. Its presence in HDL has been reported to decrease,6 or have no influence on, cholesterol efflux from cells,7,8 nor is it a cofactor for any known enzyme. Rare individuals lacking apoA-II appear to have normal plasma lipids.9 In fact, all of the known effects of apoA-II are deleterious. In studies with mice and humans, apoA-II levels have been associated with increased susceptibility to atherosclerosis,10–13 increased free fatty acid levels,14,15 increased body fat,15,16 and increased insulin resistance.15,17 Undoubtedly, apoA-II has some unknown beneficial functions, possibly related to fatty acid metabolism or host defense. In this issue of Arteriosclerosis, Thrombosis, and Vascular Biology , …