04512nam 2200745Ii 45000010011000000030004000110050017000150060024000320070015000560080041000710200030001120200025001420240028001670400030001950500025002250820016002501000035002662450159003012640056004603000056005163360021005723370026005933380032006194900086006515040057007375050023007945050140008175050163009575050093011205050082012135050141012955050093014365050104015295050090016335050091017235050127018145060065019415100019020065100023020255100011020485100011020595100039020705100031021095100019021405100024021595100026021835200708022095240238029175300029031555380036031845380047032205880046032676500039033136500025033526550022033777000027033997000035034267000034034617000029034957000044035247100032035687760034036008300057036348560075036912400000035NOW20210328190106.0m eo d cr cn |||m|||a190401s2021 maua ob 000 0 eng d a9781680837872qelectronic z9781680837865qprint7 a10.1561/24000000352doi aCaBNVSLcCaBNVSLdCaBNVSL 4aQA76.87b.L58 2021eb04a006.3/22231 aLiu, Changliu,d1990-eauthor.10aAlgorithms for verifying deep neural networks /cChangliu Liu, Tomer Arnon, Christopher Lazarus, Christopher Strong, Clark Barrett, Mykel J. Kochenderfer. 1a[Hanover, Massachusetts] :bNow Publishers,c2021. a1 PDF (pages 244-404) :billustrations (some color) atext2rdacontent aelectronic2isbdmedia aonline resource2rdacarrier1 aFoundations and trends in optimization,x2167-3918 ;vVol. 4: No. 3-4, pp 244-404 aIncludes bibliographical references (pages 397-404).0 a1. Introduction --8 a2. Problem formulation. 2.1. Feedforward neural network ; 2.2. Verification problem ; 2.3. Results ; 2.4. Soundness and completeness --8 a3. Overview of methods. 3.1. Reachability ; 3.2. Primal optimization ; 3.3. Dual optimization ; 3.4. Search and reachability ; 3.5. Search and optimization --8 a4. Preliminaries. 4.1. Bounds ; 4.2. Set split ; 4.3. Gradient ; 4.4. ReLU activation --8 a5. Reachability. 5.1. Overview ; 5.2. ExactReach ; 5.3. Ai2 ; 5.4. MaxSens --8 a6. Primal optimization. 6.1. Encoding a network as constraints ; 6.2. Objective functions ; 6.3. NSVerify ; 6.4. MIPVerify ; 6.5. ILP --8 a7. Dual optimization. 7.1. Dual network ; 7.2. Duality ; 7.3. ConvDual ; 7.4. Certify --8 a8. Search and reachability. 8.1. ReluVal ; 8.2. Neurify ; 8.3. FastLin ; 8.4. FastLip ; 8.5. DLV --8 a9. Search and optimization. 9.1. Sherlock ; 9.2. BaB ; 9.3. Planet ; 9.4. Reluplex --8 a10. Comparison and results. 10.1. Bound experiments ; 10.2. Performance experiments --8 a11. Conclusion. 11.1. Computational efficiency ; 11.2. Future directions ; 11.3. Summary -- Acknowledgments -- References. aRestricted to subscribers or individual document purchasers.0 aGoogle Scholar0 aGoogle Book Search0 aScopus0 aINSPEC0 aDBLP Computer Science Bibliography0 aZentralblatt MATH Database0 aAMS MathSciNet0 aACM Computing Guide0 aACM Computing Reviews3 aDeep neural networks are widely used for nonlinear function approximation, with applications ranging from computer vision to control. Although these networks involve the composition of simple arithmetic operations, it can be very challenging to verify whether a particular network satisfies certain input-output properties. This article surveys methods that have emerged recently for soundly verifying such properties. These methods borrow insights from reachability analysis, optimization, and search. We discuss fundamental differences and connections between existing algorithms. In addition, we provide pedagogical implementations of existing methods and compare them on a set of benchmark problems. aChangliu Liu, Tomer Arnon, Christopher Lazarus, Christopher Strong, Clark Barrett and Mykel J. Kochenderfer (2021), "Algorithms for Verifying Deep Neural Networks", Foundations and Trends in Optimization: Vol. 4: No. 3-4, pp 244-404. aAlso available in print. aMode of access: World Wide Web. aSystem requirements: Adobe Acrobat reader.0 aTitle from PDF (viewed on Mar. 28, 2021). 0aNeural networks (Computer science) 0aComputer algorithms. 0aElectronic books.1 aArnon, Tomer,eauthor.1 aLazarus, Christopher,eauthor.1 aStrong, Christopher,eauthor.1 aBarrett, Clark,eauthor.1 aKochenderfer, Mykel J.,d1980-eauthor.2 aNow Publishers,epublisher.08iPrint version:z9781680837865 0aFoundations and trends in optimization ;x2167-3918 483Abstract with links to full textuhttp://dx.doi.org/10.1561/240000003507015nam 2200637Ii 45000010011000000030004000110050017000150060024000320070015000560080041000710200030001120200025001420240028001670400030001950500023002250820019002481000032002672450051002992640056003503000024004063360021004303370026004513380032004774900085005095040057005945050023006515050646006745050296013205050227016165050222018435050480020655050301025455050275028465050163031215050046032845060065033305100019033955100023034145100011034375100011034485100031034595100019034905202292035095240124058015300029059255380036059545380047059905880046060376500018060836500028061016550022061297100032061517760034061838300085062178560075063020800000035NOW20210328190106.0m eo d cr cn |||m|||a190401s2021 mau ob 000 0 eng d a9781680837933qelectronic z9781680837926qprint7 a10.1561/08000000352doi aCaBNVSLcCaBNVSLdCaBNVSL 4aHB139b.M64 2021eb04a330.0151952231 aMoffatt, Peter G.,eauthor.10aExperimetrics :ba survey /cPeter G. Moffatt. 1a[Hanover, Massachusetts] :bNow Publishers,c2021. a1 PDF (pages 1-152) atext2rdacontent aelectronic2isbdmedia aonline resource2rdacarrier1 aFoundations and trends in econometrics,x1551-3084 ;vVol. 11: No. 1-2, pp 1-152 aIncludes bibliographical references (pages 141-152).0 a1. Introduction --8 a2. Treatment testing. 2.1. Key concepts and definitions ; 2.2. Choosing the values of "a" and "B" ; 2.3. Treatment tests illustrated ; 2.4. Dictator game data ; 2.5. Tests of normality ; 2.6. Parametric (independent-sample) treatment tests ; 2.7. Non-parametric (independent-sample) treatment tests ; 2.8. The bootstrap ; 2.9. Tests comparing entire distributions ; 2.10. Independent-sample tests with binary outcomes ; 2.11. Within-subject tests ; 2.12. Within-subject tests with binary outcomes ; 2.13. Within-subject tests with binary outcomes: other applications ; 2.14. Treatment testing using regression models and multilevel models --8 a3. Power analysis.3.1. Power analysis -- theory ; 3.2. Power analysis -- practice ; 3.3. Power and the scientific quality debate ; 3.4. The Monte Carlo method in power calculations ; 3.5. Power of treatment tests in multilevel models ; 3.6. Choosing number of subjects and number of tasks --8 a4. Experimental data types. 4.1. Binary data ; 4.2. Optimal design of binary choice problems ; 4.3. Ordinal data ; 4.4. Interval data ; 4.5. Multivariate interval data ; 4.6. Continuous (exact) data ; 4.7. Censored data --8 a5. Structural estimation of social preference parameters. 5.1. The modified dictator game ; 5.2. Estimation of social preference parameters ; 5.3. Estimation of social preference parameters using stated choice data --8 a6. Continuous heterogeneity: maximum simulated likelihood. 6.1. Theoretical background for choice under risk ; 6.2. Decision-theoretical framework: EU and RDU ; 6.3. The Fechner model (random utility model) ; 6.4. The tremble parameter ; 6.5. The role of experience ; 6.6. Between-subject variation and the sample log-likelihood ; 6.7. The method of maximum simulated likelihood (MSL) ; 6.8. Post-estimation ; 6.9. The random preference (RP) model ; 6.10. Non-nested tests --8 a7. Discrete heterogeneity: finite mixture models. 7.1. Finite mixture models ; 7.2. Depth of reasoning models ; 7.3. The 11â€“20 money request game ; 7.4. Guessing games ; 7.5. Other depth of reasoning models ; 7.6. Other applications of the finite mixture model ; 7.7. Machine learning models --8 a8. Other models for behaviour in games. 8.1. Modelling choices in repeated games ; 8.2. Non-parametric tests on repeated game data ; 8.3. Quantal response equilibrium (QRE): theory ; 8.4. Computing the probabilities in the QRE model ; 8.5. Estimation of the QRE model --8 a9. Models of learning. 9.1. Directional learning ; 9.2. Reinforcement learning ; 9.3. Belief learning ; 9.4. The experience weighted attraction (EWA) model --8 a10. Conclusion -- Appendix -- References. aRestricted to subscribers or individual document purchasers.0 aGoogle Scholar0 aGoogle Book Search0 aINSPEC0 aScopus0 aZentralblatt MATH Database0 aAMS MathSciNet3 aThis monograph aims to survey a range of econometric techniques that are currently being used by experimental economists. It is likely to be of interest both to experimental economists who are keen to expand their skill sets, and also the wider econometrics community who may be interested to learn the sort of econometric techniques that are currently being used by Experimentalists. Techniques covered range from the simple to the fairly advanced. The monograph starts with an overview of treatment testing. A range of treatment tests will be illustrated using the example of a dictator-game giving experiment in which there is a communication treatment. Standard parametric and non-parametric treatment tests, tests comparing entire distributions, and bootstrap tests will all be covered. It will then be demonstrated that treatment tests can be performed in a regression framework, and the important concept of clustering will be explained. The multilevel modelling framework will also be covered, as a means of dealing with more than one level of clustering. Power analysis will be covered from both theoretical and practical perspectives, as a means of determining the sample size required to attain a given power, and also as a means of computing ex-post power for a reported test. We then progress to a discussion of different data types arising in Experimental Economics (binary, ordinal, interval, etc.), and how to deal with them. We then consider the estimation of fully structural models, with particular attention paid to the estimation of social preference parameters from dictator game data, and risky choice models with between-subject heterogeneity in risk aversion. The method maximum simulated likelihood (MSL) is promoted as the most suitable method for estimating models with continuous heterogeneity. We then consider finite mixture models as a way of capturing discrete heterogeneity; that is, when the population of subjects divides into a small number of distinct types. The application used as an example will be the level-k model, in which subject types are defined by their levels of reasoning. We then consider other models of behaviour in games, including the Quantal Response Equilibrium (QRE) Model. The final area covered is models of learning in games. aPeter G. Moffatt (2021), "Experimetrics: A Survey", Foundations and Trends in Econometrics: Vol. 11: No. 1-2, pp 1-152. aAlso available in print. aMode of access: World Wide Web. aSystem requirements: Adobe Acrobat reader.0 aTitle from PDF (viewed on Mar. 28, 2021). 0aEconometrics. 0aExperimental economics. 0aElectronic books.2 aNow Publishers,epublisher.08iPrint version:z9781680837926 0aFoundations and trends in econometrics ;vVol. 11: No. 1-2, pp 1-152.x1551-3084483Abstract with links to full textuhttp://dx.doi.org/10.1561/080000003506192nam 2200661Ii 45000010011000000030004000110050017000150060024000320070015000560080041000710200030001120200025001420240028001670400030001950500029002250820016002541000034002702450112003042640056004163000056004723360021005283370026005493380032005754900092006075040057006995050184007565050342009405050272012825050272015545050207018265050190020335050340022235050116025635060065026795100019027445100023027635100011027865100011027975100024028085100026028325100039028585100031028975100019029285201892029475240197048395300029050365380036050655380047051015880046051486500018051946550022052127000030052347000032052647100032052967760034053288300093053628560075054552000000109NOW20210328190106.0m eo d cr cn |||m|||a190401s2021 maua ob 000 0 eng d a9781680837919qelectronic z9781680837902qprint7 a10.1561/20000001092doi aCaBNVSLcCaBNVSLdCaBNVSL 4aTK5103.4836b.D46 2021eb04a621.3842231 aDemir, čOzlem Tugfe,eauthor.10aFoundations of user-centric cell-free massive MIMO /cčOzlem Tugfe Demir, Emil Bjčornson, Luca Sanguinetti. 1a[Hanover, Massachusetts] :bNow Publishers,c2021. a1 PDF (pages 162-472) :billustrations (some color) atext2rdacontent aelectronic2isbdmedia aonline resource2rdacarrier1 aFoundations and trends in signal processing,x1932-8354 ;vVol. 14: No. 3-4, pp 162-472 aIncludes bibliographical references (pages 451-472).0 a1. Introduction and motivation. 1.1. Cell-free networks ; 1.2. Historical background ; 1.3. Three benefits over cellular networks ; 1.4. Summary of the key points in section 1 -- 8 a2. User-centric cell-free massive MIMO networks. 2.1. Definition of cell-free massive MIMO ; 2.2. User-centric dynamic cooperation clustering ; 2.3. System models for uplink and downlink ; 2.4. Network scalability ; 2.5. Channel modeling ; 2.6. Channel hardening and favorable propagation ; 2.7. Summary of the key points in section 2 --8 a3. Theoretical foundations. 3.1. Estimation theory for Gaussian variables ; 3.2. Capacity bounds and spectral efficiency ; 3.3. Maximization of Rayleigh quotients ; 3.4. Optimization algorithms for utility maximization ; 3.5. Summary of the key points in section 3 --8 a4. Channel estimation. 4.1. Uplink pilot transmission ; 4.2. MMSE channel estimation ; 4.3. Impact of architecture, contamination, & spatial correlation ; 4.4. Pilot assignment and dynamic cooperation cluster formation ; 4.5. Summary of the key points in section 4 --8 a5. Uplink operation. 5.1. Centralized uplink operation ; 5.2. Distributed uplink operation ; 5.3. Running example ; 5.4. Numerical performance evaluation ; 5.5. Summary of the key points in section 5 --8 a6. Downlink operation. 6.1. Centralized downlink operation ; 6.2. Distributed downlink operation ; 6.3. Numerical performance evaluation ; 6.4. Summary of the key points in section 6 --8 a7. Spatial resource allocation. 7.1. Transmit power optimization ; 7.2. Scalable distributed power optimization ; 7.3. Comparison of power optimization schemes ; 7.4. Pilot assignment ; 7.5. Selection of dynamic cooperation clusters ; 7.6. Implementation constraints ; 7.7. Summary of the key points in section 7 -- Acknowledgements --8 aAppendices. C.1. Proofs from section 4 ; C.2. Proofs from section 5 ; C.3. Proofs from section 6 -- References. aRestricted to subscribers or individual document purchasers.0 aGoogle Scholar0 aGoogle Book Search0 aINSPEC0 aScopus0 aACM Computing Guide0 aACM Computing Reviews0 aDBLP Computer Science Bibliography0 aZentralblatt MATH Database0 aAMS MathSciNet3 aImagine a coverage area where each mobile device is communicating with a preferred set of wireless access points (among many) that are selected based on its needs and cooperate to jointly serve it, instead of creating autonomous cells. This effectively leads to a user-centric post-cellular network architecture, which can resolve many of the interference issues and service-quality variations that appear in cellular networks. This concept is called User-centric Cellfree Massive MIMO (multiple-input multiple-output) and has its roots in the intersection between three technology components: Massive MIMO, coordinated multipoint processing, and ultra-dense networks. The main challenge is to achieve the benefits of cell-free operation in a practically feasible way, with computational complexity and fronthaul requirements that are scalable to enable massively large networks with many mobile devices. This monograph covers the foundations of User-centric Cell-free Massive MIMO, starting from the motivation and mathematical definition. It continues by describing the state-of-the-art signal processing algorithms for channel estimation, uplink data reception, and downlink data transmission with either centralized or distributed implementation. The achievable spectral efficiency is mathematically derived and evaluated numerically using a running example that exposes the impact of various system parameters and algorithmic choices. The fundamental tradeoffs between communication performance, computational complexity, and fronthaul signaling requirements are thoroughly analyzed. Finally, the basic algorithms for pilot assignment, dynamic cooperation cluster formation, and power optimization are provided, while open problems related to these and other resource allocation problems are reviewed. All the numerical examples can be reproduced using the accompanying Matlab code. ačOzlem Tugfe Demir, Emil Bjčornson and Luca Sanguinetti (2021), "Foundations of User-Centric Cell-Free Massive MIMO", Foundations and Trends in Signal Processing: Vol. 14: No. 3-4, pp 162-472. aAlso available in print. aMode of access: World Wide Web. aSystem requirements: Adobe Acrobat reader.0 aTitle from PDF (viewed on Mar. 28, 2021). 0aMIMO systems. 0aElectronic books.1 aBjčornson, Emil,eauthor.1 aSanguinetti, Luca,eauthor.2 aNow Publishers,epublisher.08iPrint version:z9781680837902 0aFoundations and trends in signal processing ;vVol. 14: No. 3-4, pp 162-472.x1932-8354 483Abstract with links to full textuhttp://dx.doi.org/10.1561/2000000109