Embedding ASP in mobile systems: discussion and preliminary implementation

Francesco Calimeri, Davide Fuscà, Stefano Germano, Simona Perri, and Jessica Zangari

Department of Mathematics and Computer Science, University of Calabria, Italy {calimeri,fusca,germano,perri,zangari}@mat.unical.it

Answer Set Programming (ASP) is a well-established declarative problem solving paradigm which became widely used in AI and recognized as a powerful tool for knowledge representation and reasoning (KRR), especially for its high expressiveness and the ability to deal also with incomplete knowledge. Recently, thanks to the availability of a number of robust and ecient implementations, ASP has been increasingly employed in a number of dierent domains, and used for the development of industrial-level and enterprise applications. Notably, this has been fostered by the release of a variety of proper development tools and interoperability mechanisms for allowing interaction and integration with external systems; however, there is still a lack of proper means for an eective applicability of ASP in the mobile setting. In this preliminary work, we present a novel general framework for embedding ASP in external systems, and in particular within applications for mobile devices. We show the use of the framework by illustrating the integration of the ASP system DLV into Android applications, and the development of , a customizable health app that helps the user to meet her daily tness goals, as a proof of concept. Abstract.

DLVfit

Keywords:

Answer Set Programming, DLV, Android, Mobile

1 Introduction Answer Set Programming (ASP) [4,5,11,12,17,23,24] is a powerful declarative formalism for knowledge representation and reasoning developed in the eld of logic programming and nonmonotonic reasoning. After more than twenty years of scientic research, the theoretical properties of ASP are well understood and the solving technology, as witnessed by the availability of a number of robust and ecient systems [7], is mature for practical applications; indeed, ASP has been increasingly employed in a number of dierent domains, and used for the development of industrial-level and enterprise applications [8,22]. Notably, this has been fostered by the release of a variety of proper development tools and interoperability mechanisms for allowing interaction and integration with external systems [6,14,15,16]. However, the worldwide commercial and industrial scenario

signicantly changed in the latest years; smartphones, or smart/wearable devices in general, are constantly gaining popularity as computational power and features increase, in terms of sensors, communication means and applications availability. In this context, that forced the whole ICT industry to radically change, there is no work in literature about ASP systems that can natively run on mobile devices, to our knowledge; this leaves the potential of ASP still undisclosed in this scenario. Indeed, having a native ASP solver along with proper development tools for the integration of ASP modules within standard software products, would pave the way to a new generation of mobile applications, that might signicantly take advantage from the knowledge representation and reasoning capabilities of ASP. At rst, this could be of great help for the ASP developers that, beside porting already existing ASP-based applications to the mobile world, can also think of completely new scenarios for the fruitful application of ASP, both in research and industry. In this preliminary work, we introduce embASP, a novel framework for the integration of ASP in external systems; notably, it can be easily extended in order to match any platform and ASP solver through plain Java Object-Oriented programming. In order to show the potential of embASP, we present here an actual implementation of the framework for the Android platform [1] and the

DLV

system [21], thus eectively bringing ASP on mobile systems. Furthermore,

as a proof of concept, we present DLVfit, an ASP-based mobile health app designed and implemented by means of the embASP framework, thus resulting easy to use and customize. The framework, along with a beta version of DLVfit and the details about the logic program therein employed, is freely available online [13].

2 The embASP Framework We start by illustrating the abstract architecture of embASP, and then provide the reader with some insights about a specialization designed for making use of the

2.1

DLV

system within Android applications.

Abstract Architecture

The framework consists of three dierent layers, and Figure 1 shows how they interact:



the rst layer, ASPHandler, acts like a façade to the user of the framework: it contains all functions needed in order to manage input and output of the



ASP solver; the second layer, Solver Handler, acts as a middleware between the façade and the actual solver, providing the native functions that allow the invocation



of the solver; the third layer, ASPSolver, concerns the actual solver and contains the functions to run and make use of it.

ASPHANDLER

SOLVER HANDLER Android Service Java Native Interface (JNI)

ASP SOLVER DLV

Fig. 1: Abstract Architecture of the embASP framework. Overshadowed components are related to the specialization for

2.2

DLV

on Android.

Framework implementation

The abstract architecture (Figure 1) has been kept simple, and conceived in order to t any Java compliant environment, as well as any kind of ASP solvers. Basically, the execution works as follows: given a logic program

P,

the AS-

PHandler (asynchronously) calls the Solver Handler, providing it with

P,

along with the proper options for the solver at hand; then, the Solver Handler starts the ASPSolver by invoking the native functions of the solver. Eventually, the results are published and turned back to the caller via a callback function by the ASPHandler. The ASPSolver features a general interface, while within its implementation specic calls to the particular solvers it is specialized for. Notably, the framework allows to dynamically change at runtime the underlying logic program (from the addition of new rules, to the activation of dierent optimization statements, to the update of the facts base, towards a brand new program), and also to change solver options; this enables changes of behaviour while the application is running. Figure 2 shows how the layers are eectively implemented in embASP: the

ASPService and ASPHandler, along with their subclasses, represent the Solver Handler and the ASPHandler layers, respectively, while the dierent solvers

compose the ASPSolver layer. Notably, if needed, one can raise from the use of a single ASP system to the management of multiple instances of dierent solvers. The abstract class

ASPHandler

provides all means for preparing input

for the solvers, in several ways (simple strings, les, Java Objects); in addition, it manages the setting of all options for the actual ASP solvers; it also features proper methods for making the reasoning start. The abstract class

ASPService

Client

ASP HANDLER ASPHandler

Mapper

AnswerSetCallback

AnswerSet ASPHandler_Solver1

...

AnswerSets

ASPHandler_SolverN

AnswerSets_Solver1

...

AnswerSets_SolverN

SOLVER HANDLER ASPService

ASPService_Solver1

...

ASPService_SolverN

ASP SOLVER Solver1

SolverN

Fig. 2: embASP: class diagram.

is in charge of managing invocations to the actual ASP solvers, and gathering the results. Each solver invocation is asynchronous; thus, once a reasoning task has produced the answer set(s), a callback function is automatically red. It can be specied by implementing the

AnswerSetCallback interface, and all answer sets AnswerSets class.

are captured and parsed by the The framework features a

mapper to convert the output from the solvers into

Java Objects: the process is guided by proper annotations over involved classes. In order to map a class to a predicate, there are two types of annotations:

@Predicate(string_name), for the predicate name (the target must be a class), and @Term(integer_position), for the term and its position in the atom (the target must be a eld of a class annotated via @Predicate ). Currently, the data types supported by the ASP solvers are strings and integers; on the Java side, yet allowing for more advanced integration with Java objects, embASP preserves the mapping; however, whenever the ASP data types will be changed, the framework design will allow it to be easily extended accordingly. The mapping works also the other way around, i.e., from objects to strings, so that the input can be also specied as plain Java Objects. The mapper is meant to help the developer at dividing an application into two separated modules: an ASP reasoning module and a Java module. Indeed, the mapper acts like a middle-ware that enables the communication among the

modules, and facilitates the developer's job by providing her with an explicit and ready-made mapping between Java objects and the logic part, without the need for any translation from string to objects and vice-versa. Such separation helps at keeping things simple when developing complex application: think about a scenario in which dierent gures are involved, such as Android/Java developers and KRR experts; both gures can benet from the fact that the knowledge base and the reasoning modules can be designed and developed independently from the rest of the Java-based application. We note here that, as already outlined above, in the spirit of generality and usability, the framework allows the developer to design and build applications that make use of multiple concurrent instances of an ASP solver; in addition, if needed, dierent specializations of the ASPSolver layer can be used at once, thus making possible the actual concurrent run of dierent ASP solvers. Figure 2 shows an example of how the framework can be extended for multiple solvers in order to t the specic needs of the users. The

ASPHandler

class must be

extended, and each concrete class extending it should start a proper

ASPService.

Moreover, in order to deal with the specic output of the solver at hand, the

AnswerSets class must be specialized as well. In general, the number of dierent solvers that the framework can manage is only limited by the computational resources featured by the platform/device the application will be executed.

3 An embASP Specialization for Android and

DLV

As already introduced above, we present here a specialization of embASP explicitly tailored for the development of native Android applications making use of ASP by means of the

DLV

system.

Android is by far the most used mobile operating system worldwide; the ocial version has been originally released, and is currently developed, by Google, and its popularity is due also to an open source nature, that allows any manufacturer to customize and adapt it to an incredibly wide range of devices, not just smartphones. The Android development model is based on Java (indeed, running Android apps are managed by a special Java Virtual Machine, currently called

ART [2]), and Google provides the developers with a large set of tools

and libraries.

DLV

[21] is the rst ASP system which is undergoing an industrial

exploitation by a spin-o company called DLVSystem Ltd. [9], thus fostering the interest of several industries in ASP and

DLV

itself [8,19,22].

We provide next some details about the specialization of the embASP framework for

DLV

and Android, and some technical aspects concerning the actual

porting of such solver on the Android platform.

3.1

The

embASP Specialization

The additional components w.r.t. the abstract architecture needed to implement the specialization are depicted as overshadowed in Figure 1; Figure 3 reports the resulting hierarchy of classes.

Client

ASPHandler Lstart,asCallbackP:PAnswerSetCallbackv LaddOption,optionsP:PStringv LaddRawInput,programP:PStringv LaddFileInput,filePathP:PStringv LaddInput,objP:PObjectv LaddInput,objectsP:PSetvPSet LaddFilter,objP:PClassv LaddFilter,predicateP:PStringv

ASPMapper LregisterClass,classP:PClassv LgetClass,predicateP:PStringvP:PClass LgetString,objP:PObjectvP:PString LgetObject,stringP:PStringvP:PObject

AnswerSetCallback Lcallback,answerSetsP:PSetv

AnswerSets -answerSetListP:PList Lparse,v AnswerSet -outputStringP:Pstring -weightMapP:Pmap

DLVHandler Lstart,asCallbackP:PAnswerSetCallbackv LaddFilter,predicateP:PStringv LaddFilter,objP:PClassv

LgetAnswerObjects,vP:PSet

DLVAnswerSets Lparse,v

ASPService LhandleActionSolve,v

DLVService LhandleActionSolve,v

Fig. 3: embASP: class diagram as specialized for

The

DLVHandler

and

DLVAnswerSets

DLV.

classes relate to the ASPHandler

layer and are the specialization of the respective abstract classes designed for

DLV.

These classes provide a proper implementation tailored for the specic

functionalities of the solver, i.e. the setting of its peculiar options, and the parsing of its output format. The

DLVService class is related to the Solver Handler ASPService class for the Android underlying

cialization of the

enables the invocation of

DLV

layer as a spetechnologies. It

by means of Android Services, that are native ap-

plication components able to perform long-running operations in the background. They allow the applications that make use of embASP to asynchronously execute other tasks while waiting for the answers from the ASP solver; furthermore, the use of Android services makes the solvers type and invocation details completely transparent to the rest of the Android app. The specialization of the ASPSolver layer to

DLV required an explicit port-

ing, since it is natively developed in C++; in particular, we used JNI (Java Native Interface) [20] in order to let the upper layers communicate with the

ASPSolver one.

3.2

The actual porting

For the actual porting of is based on

DLV

on Android, whose standard development process

Java, we made use of the Android NDK (Native Development Kit) [3],

that allows developers to implement parts of an Android application as native-

code languages, such as C and C++, and it is well-suited for CPU-intensive workloads such as game engines, signal processing, physics simulation, and so on. Moreover, the use of JNI grants the access to the API provided by the Android NDK, and therefore the access to the exposed

DLV

functionalities directly from

the Java code of an Android application. It is worth noting that

DLV

is a very

complex system, therefore, in order to successfully complete the porting, an equally elaborated setup did not suce, and a subtle and extended work of adjustments within the code had to be carried out. In order to provide the reader with an idea, we can report here that some of the major actions were related to a re-design of the structure of the

DLV

project (starting from makeles) in

order to make it NDK compliant, and a deep revision of the

DLV code in order to

make it executable as an Android Service, taking into account its rather complex nature due to the well-known algorithms and techniques

DLV

relies on (see [21]

and related literature) and to several advanced optimizations, many of them explicitly relying on the compiler; we do think that further technicalities would be of little or no interest to the reader and the sake of the paper contribution.

4 A rst full native ASP-based Android App: DLVfit As a proof of concept, in order to show the framework features and capabilities, in this Section we introduce DLVfit, a simple Android health app that aims at suggesting the owner of a mobile device the best way, according to user's desiderata, to achieve some tness goals. The app lets the user express her own goals and preferences in a very customizable ways along many composable dimensions: calories to burn, time to spend, dierentiation over several physical activities, time constraints, etc. Then, it monitors her actual activity throughout the day and, upon request, it computes one or more plans meant, if accomplished, to make her meet the aforementioned goals the way she would have preferred. More in detail, the app constantly detects the current user activity (running, walking, cycling, etc.) and (at a customizable frequency) stores some information (activity type, timestamps, calories burned up to the present time, etc.). Activity detection is performed by means of the Google Activity Recognition APIs [18], a de-facto standard on Android, thus relying on these for the accuracy of detection. As already mentioned, the user might ask, at any time, for a suggestion about a plan for the rest of the day; the reasoning module hence prepares a (set of ) proper workout plans complying with the very personal goals and preferences previously expressed. The reasoning module is actually in charge of building a proper ASP program, which is in turn fed to

DLV

via embASP. Such program matches the classical

guess/check/optimize paradigm, thus resulting easy to understand, enrich and customize.

 

The guess part chooses how much time to spend on each exercise; the check forces the resulting plan to be admissible: burning the remaining amount of desired calories, do not exceed the time constraints, etc.;

(a)

(b)

Fig. 4: Screenshots from DLVfit app: main menu (a) and list of optimizations (b).



the optimize part, eventually, expresses preferences: minimize total time spent exercising, number of activities to perform, maximize the number of dierent activity types, avoid activities around a given time of the day, etc.

There is a wide range of customization possibilities in this setting: thanks to the modeling capabilities and the declarative nature of ASP, adding new features to

DLVfit, such as new exercises or new kind of preferences, is straightforward, and sums up to adding a few lines to the logic program. It is also worth noting that the ASP program is dynamically built, thus providing the developer (and, in turn, the nal user) with great customization and exibility capabilities. We remind here that DLVfit is meant to be a proof of concept; it can be signicantly improved, and must be considered as a beta. However, it is fully functional, and is freely available online, along with further details, the full version of the base ASP program therein employed, a detailed documentation and an embASP tutorial [13].

5 Related Work The problem of embedding ASP reasoning modules into external systems and/or externally controlling an ASP system has been already investigated in the lit-

erature; to our knowledge, the more widespread solutions are the

DLV

Java

Wrapper [25], JDLV [14], and the scripting facilities featured by clingo4 [16], which allow, to dierent extents, the interaction and the control of ASP solvers from external applications. In clingo4, the scripting languages

lua and python enable a form of control

over the computational tasks of the embedded solver clingo, with the main purpose of supporting also dynamic and incremental reasoning; on the other hand,

embASP, similarly to the Java Wrapper and JDLV, acts like a versatile wrapper wherewith the developers can interact with the solver. However, dierently from the Java Wrapper, embASP makes use of Java Annotations, a form of metadata that can be examined at runtime, thus allowing an easy mapping of input/output to Java Objects; and dierently from JDLV, that uses JPA annotations for dening how Java classes map to relations similarly to ORM frameworks, embASP straightforwardly uses custom annotations, almost eortless to dene, to deal with the mapping. Moreover, our framework is not specically bound to a single or specic solver; rather, it can be easily extended for dealing with dierent solvers, as shown in Figure 2; in addition, it allows to build applications that can run dierent solvers, and dierent instances, at the same time; none of the mentioned systems exposes this feature. Finally, to our knowledge, the specialization of

embASP for

DLV

on Android is the rst actual attempt to port ASP solvers to

mobile systems reported in literature. Several ways of taking advantage from ASP capabilities have been explored, and, interestingly, not all of them require to natively port an ASP system on the device of use. In particular, it is possible to let the reasoning tasks take place somewhere else, and use internet connections in order to communicate between the reasoning service and the actual application, according to a cloud computing paradigm, to some extent. Thanks to such mechanisms, mobile apps relying on ASP reasoning have already been introduced: in [10] a prototype system is presented, called

HealthyLife, which makes use of ASP-based Stream Reasoning

(ASR) in a mobile health app that has some point of contacts with DLVfit. The focus of

HealthyLife is primarily to detect users daily activities and try to deal

with ambiguities when recognizing situations, while DLVfit delegates this task to Android Recognition API: its primary goal is to experiment with the usage of ASP on mobile devices. In this respect, although the computational power of a dedicated server is not comparable to the one of a mobile device, it would be interesting to see whether

HealthyLife could benet from the embedding of DLV

and embASP within it. The fact that

HealthyLife works on a cloud-based manner, while, thanks to

embASP, DLVfit natively embeds an ASP solver, deserves some considerations. Provided that this is not intended as a discussion on the matter, there are clear dierences between the two approaches. The cloud-based approach grants great computational power to low-end devices, without the need for actually porting a system to the nal user's device, and completely preventing any performance issue. However, in order this to take place, there's rst the need for a proper

application hosting, which requires non-negligible eorts both from the design and the economic points of view; furthermore, a steady internet connection might be a strong constraint, especially when the communication between the end user's device and the cloud infrastructure requires a large bandwidth. On the other hand, a native-based approach might involve signicant eorts for the actual porting of pieces of softwares on the target device, which, in turn, might lead to performance or power consumption issues; and even if performance issues might not appear as always crucial, given the computational power which is available even on mobile devices, power consumption is sometimes decisive. Nevertheless, in our showcase scenario, DLVfit shows that the development of applications that natively runs ASP-based reasoning tasks on mobile devices does not necessarily suer from the discussed drawbacks. Indeed,

DLV is invoked

only on demand, i.e., whenever the user wants to check possible alternatives about how to spend the rest of her day; for the whole rest of the time, no solver is running or waiting, thus preventing both performance and battery drain.

6 Conclusions In this seminal work we introduced embASP, a general framework for embedding the reasoning capabilities of ASP into external systems; in order to assess the framework capabilities, we presented also a specialization for the mobile setting that is tailored for making use of

DLV

within Android apps. In addition, we

presented a preliminary version of DLVfit, an Android health app that shows the eectiveness of the framework and represents, to our knowledge, the rst mobile app natively running an ASP system. The framework, the app and further details are freely available online [13]. As future work, we plan to test the framework over dierent platforms and extend it to dierent ASP solvers; moreover, apart from rening DLVfit, we want to further investigate the potential of ASP on mobile systems by means of new applications.

7 Acknowledgements The authors would like to thank Giovambattista Ianni for all suggestions, constructive comments and fruitful discussions.

References Android, http://www.android.com Android ART, https://source.android.com/devices/tech/dalvik/ Android NDK, https://developer.android.com/tools/sdk/ndk Baral, C.: Knowledge Representation, Reasoning and Declarative Problem Solving. Cambridge University Press (2003) 5. Brewka, G., Eiter, T., Truszczynski, M.: Answer set programming at a glance. Commun. ACM 54(12), 92103 (2011) 1. 2. 3. 4.

6. Calimeri, F., Cozza, S., Ianni, G.: External sources of knowledge and value invention in logic programming. Annals of Mathematics and Articial Intelligence 50(34), 333361 (2007) 7. Calimeri, F., Gebser, M., Maratea, M., Ricca, F.: The design of the fth answer set programming competition. CoRR abs/1405.3710 (2014) 8. Calimeri, F., Ricca, F.: On the application of the answer set programming system dlv in industry: a report from the eld. Book Reviews 2013(03) (2013) 9. DLVSystem Ltd., http://www.dlvsystem.com 10. Do, T.M., Loke, S.W., Liu, F.: Healthylife: An activity recognition system with smartphone using logic-based stream reasoning. In: Mobile and Ubiquitous Systems: Computing, Networking, and Services, pp. 188199. Springer (2013) 11. Eiter, T., Faber, W., Leone, N., Pfeifer, G.: Declarative problem-solving using the dlv system. In: Logic-based articial intelligence, pp. 79103. Springer (2000) 12. Eiter, T., Ianni, G., Krennwallner, T.: Answer Set Programming: A Primer. In: Reasoning Web. Semantic Technologies for Information Systems, 5th International Summer School - Tutorial Lectures. pp. 40110. Brixen-Bressanone, Italy (AugustSeptember 2009) 13. , https://www.mat.unical.it/calimeri/projects/embasp/ 14. Febbraro, O., Grasso, G., Leone, N., Ricca, F.: JASP: a framework for integrating Answer Set Programming with Java. In: Proc. of KR2012. AAAI Press (2012) 15. Febbraro, O., Reale, K., Ricca, F.: ASPIDE: Integrated Development Environment for Answer Set Programming. In: Proc. of LPNMR'11. Springer Verlag (2011) 16. Gebser, M., Kaminski, R., Kaufmann, B., Schaub, T.: Clingo= asp+ control: Preliminary report. arXiv preprint arXiv:1405.3694 (2014) 17. Gelfond, M., Lifschitz, V.: Classical Negation in Logic Programs and Disjunctive Databases. New Generation Computing 9, 365385 (1991) 18. Google Activity Recognition API, https://developer.android.com/reference/ com/google/android/gms/location/ActivityRecognition.html 19. Grasso, G., Leone, N., Manna, M., Ricca, F.: Logic Programming, Knowledge Representation, and Nonmonotonic Reasoning: Essays in Honor of Michael Gelfond, Lecture Notes in AI (LNAI), vol. 6565. Springer Verlag (2011) 20. JNI, http://docs.oracle.com/javase/8/docs/technotes/guides/jni 21. Leone, N., Pfeifer, G., Faber, W., Eiter, T., Gottlob, G., Perri, S., Scarcello, F.: The DLV System for Knowledge Representation and Reasoning. ACM Transactions on Computational Logic 7(3), 499562 (Jul 2006) 22. Leone, N., Ricca, F.: Answer set programming: A tour from the basics to advanced development tools and industrial applications. In: RR2015, to appear. LNCS (2015) 23. Marek, V.W., Truszczy«ski, M.: Stable Models and an Alternative Logic Programming Paradigm. In: Apt, K.R., Marek, V.W., Truszczy«ski, M., Warren, D.S. (eds.) The Logic Programming Paradigm  A 25-Year Perspective, pp. 375398. Springer Verlag (1999) 24. Niemelä, I.: Logic Programming with Stable Model Semantics as Constraint Programming Paradigm. Annals of Mathematics and Articial Intelligence 25(34), 241273 (1999) 25. Ricca, F.: The dlv java wrapper. In: APPIA-GULP-PRODE. pp. 263274. Citeseer (2003)

embASP

Embedding ASP in mobile systems: discussion and ...

mapping; however, whenever the ASP data types will be changed, the ... and its popularity is due also to an open source nature, that allows any manufac-.
Download PDF
2MB Sizes 1 Downloads 149 Views
Report

Recommend Documents

The Mobile Bandwagon Discussion Summary.pdf
Whoops! There was a problem loading more pages. The Mobile Bandwagon Discussion Summary.pdf. The Mobile Bandwagon Discussion Summary.pdf. Open.
Software Radio in Mobile Communication Systems
1. The GSM network can be divided into three major parts. The Mobile Station. (hand-held ... A subscriber with an appropriate SIM can use the network services.
Mobile devices and systems -
Windows 7, Mac OS 4 and Android. .... Latest version of Android is Android 2.2 (Froyo) which is based on Linux Kernel 2.6.32 and ... Adobe flash 10.1 support.
Mobile devices and systems -
Windows 7, Mac OS 4 and Android. 2.4 haNdheld deviCes. Not only has there been a transformation of titanic proportions in computing devices, but.
ASP-1
(APPLICATION ORIENTED COURSE). ASP-1 : SECRETARIAL PRACTICE. Time : 2 hours. Maximum Marks : 50. (VVeightage : 70%) ... As the General Manager of Manica Steel. 10. Enterprise, write a letter to the Regional Manager of State Bank of India to open an E
Embedding Multiple Trajectories in Simulated ...
Oct 21, 2009 - the number of embedded trajectories increases so does net- ...... Song S, Miller KD, Abbott LF (2000) Competitive Hebbian learning through.
Efficient Tag Identification in Mobile RFID Systems - Semantic Scholar
model, we propose efficient solutions to identify moving RFID tags, according to the .... the amount of energy reflected by an object is dependent on the reflective area ...... an alternative scheme to estimate the number of tags. Our methodology ...
Efficient Tag Identification in Mobile RFID Systems
State Key Laboratory of Novel Software Technology,. Department of Computer .... which contains the uplink frequency and data encoding, the Q parameter ...... [18] S. R. Jeffery, M. Garofalakis, and M. J. Franklin, “Adaptive cleaning for rfid data .
Multilinear Graph Embedding: Representation and ...
This work was supported by the National Science Council of R.O.C. under contracts ... Department of Computer Science, National Tsing Hua University, Hsinchu,. Taiwan, R.O.C. (e-mail: ...... Ph.D. degree in computer science and information.
The Hardness of Embedding Grids and Walls
Yijia Chen. School of Computer Science. Fudan University ... p-EMB(K) is W[1]-hard for the classes K of all grids and all walls. See Section 2 and in particular ...
ASP Theory.pdf
Honours)!Psychology!(Erstwhile!FYUP). Semester:' ... Applied'Social'Psychology'. Section'A. Week!1 Diversity. Cultural!Diversity ... ASP Theory.pdf. ASP Theory.
Ordinal Embedding: Approximation Algorithms and ...
‡AT&T Labs — Research, [email protected] .... and distortion in “standard” embedding, which we call “metric embedding” for distinction. A contractive ...
Affinity Weighted Embedding - JMLR Workshop and Conference ...
39.2%. 5.2. ImageNet. ImageNet (Deng et al., 2009) is a large scale image dataset organized according to WordNet (Fellbaum, 1998). Con- cepts in WordNet ...
Ordinal Embedding: Approximation Algorithms and ...
the best approximation factor for unweighted graphs into the line is O(n1/2), and even for ... ‡AT&T Labs — Research, [email protected] ..... In constructing H from G, we add an edge between two vertices u1 and u2 if and only if u1 ...
results and discussion
Apr 2, 2009 - The tourism boom started in the early 1980's when a phenomenal number of Europeans flocked to this mountain hideaway. The people soon got used to the .... (DTI) and accredited by Department of Tourism (DOT). This study came up with a pr
Affinity Weighted Embedding
Jan 17, 2013 - contain no nonlinearities (other than in the feature representation in x and y) they can be limited in their ability to fit large complex datasets, and ...
results and discussion
3 Group/network and participation . .... The individual provides a service to others, or acts for the benefit of others at a personal cost. They do this in the general ...
results and discussion
DOCUMENTATION OF THE CARNATION PRODUCTION AND MARKETING OF. MOUNTAIN BLOOM ..... advantages of this production and marketing system and they could also adopt it in their own locality. 11 ..... From the sales proceeds, the cooperative deducts the tran
results and discussion
Apr 2, 2009 - like the Department of Trade and Industry, and the Department of Tourism should be facilitated by .... (DTI) and accredited by Department of Tourism (DOT). This study ..... Sagada authorities introduced some tourist programs.
results and discussion
This thesis, titled “SOCIAL CAPITAL AMONG MEMBERS OF LEPANTO ..... employees of the Lepanto Consolidated Mining Company (LCMCO). With the ..... 3.26. Neutral. Coop Collector. 50. 3.38. Much. Audit Committee. 50. 3.3. Neutral. Credit Committee. 50.
results and discussion
advantages of this production and marketing system and they could also .... found that a pre-treatment with Ag No3 + sodium Thiosulfate + sucrose gave the best.