Learning
Sections:
|
CAAM 554: Convex Optimization, Fall 2008
simplex method interior-point methods #3965 ([patch] Slovenian Django javascript translation) - Django ...
code.djangoproject.com/ticket/3965 Apr 8, 2007 ... Add/Change #3965 ([patch] Slovenian Django javascript translation). Your email or username:. Comment (you may use WikiFormatting here): ...
Nan's Games & Comics Too
Gamestop
Phone: (713) 663-6900 Toys R US
Phone: (713) 769-8697 Meet Django
Programming RubyThe Pragmatic Programmer's GuideFirst Edition (*)some books:
paper reading:
Major Breakthrough: Music's 'DNA' Decoded | Listening Post from Wired.com
blog.wired.com/music/2008/08/major-breakthro.html http://www.businessweek.com/globalbiz/content/aug2008/gb2008088_236082.htm
What Will the Olympics Do for China?Potential problems range from too much pollution to an act of terrorism. But a successful Olympics could improve Beijing's standing in the worldEdward O. Thorp - Wikipedia, the free encyclopedia
en.wikipedia.org/wiki/Edward_O._Thorp#Applied_rese... It is also worth noting that Thorp's blackjack research is one of the very few examples where results from such reached the public first, completely bypassing the usual academic journal publications and peer review process cycle. Moreover, Thorp became one of the very few, if any, applied mathematicians who risked physical harm in verifying a computer simulation. He has also stated that he considered the whole experiment an academic exercise.[4] This also marked the first time in the history of computing that a computer was used as a gambling aid.
Posted by
timothy
on Tuesday July 29, @02:43PM from the hoping-for-even-cheaper-subnotebooks dept.
Posted by
kdawson
on Tuesday July 29, @01:50AM
from the unblinking-eye dept. 
Repton writes with news of a company, Behavioral Recognition Systems, that has received 16 patents on a new video surveillance application that can convert video images into machine-readable language, and then analyze them for anomalies that suggest suspicious behavior in the camera's field of view. The software can 'recognize' up to 300 objects and establish a baseline of activity. It should go on sale in September. "...the BRS Labs technology will likely create a fair number of false positives, [the CEO] concedes. 'We think a three-to-one ratio of alerts to actual events is what the market will accept,' he says. 'We could be wrong.'"
Posted by
kdawson
on Monday July 28, @11:05PM
from the don't-stamp-so-hard dept. 
Iddo Genuth writes "Alaskan state officials have recently announced their intention to begin funding the exploration and surveying of Alaska's largest volcanoes in hopes of utilizing these as a
source of geothermal energy. They say this volcano could provide enough energy to power thousands of households, and according to some estimates, Alaska's volcanoes and hot springs could supply up to 25% of the state's energy needs."
paper reading: Algorithms for Low-Latency Remote File Synchronization
set reconciliation http://technet.microsoft.com/en-us/library/ms187103.aspx http://en.wikipedia.org/wiki/De_Bruijn_graph paper reading: a tutorial on decomposition methods for network utility maximization
mobile operating systems
http://www.reghardware.co.uk/2008/06/05/smartphone_os_linux/
http://www.limofoundation.org/what-is-the-platform.html http://code.google.com/android/ http://www.osimworld.com/newt/l/handsetsvision/osim08/ http://www.gizmos2go.com/hp-ipaq-hw6515-mobile-messenger-fa639aaba.html http://www.linuxdevices.com/products/PD7399900675.html http://en.wikipedia.org/wiki/Access_Linux_Platform operating system
UNIX
CMS BSD Virtual Memory page referenced bit LRU SWS PR-bit, valid bit, two context switches Multics Clock policy two hand policy What space is the memory management process running? Mach http://en.wikipedia.org/wiki/Mach_kernel multi-processor support Threads vs copy-on-write Distributed Shared Memory Munin http://citeseer.ist.psu.edu/5317.html eager release TreadMarks http://www.cs.rice.edu/~willy/TreadMarks/overview.html lazy release Midway http://citeseer.ist.psu.edu/bershad91midway.html ?? ping pong effect ??invalidation protocal Cache replacement policy 2Q? ARC Spin ??collocation modula 3 why can spin overcome the speed issue of microkernel os? Nooks PCC SAFE KERNEL EXTENSIONS WITHOUT RUN-TIME CHECKING formal safety proof proof validation http://raw.cs.berkeley.edu/pcc.html#experiments http://en.wikipedia.org/wiki/First-order_logic http://en.wikipedia.org/wiki/LF_(logical_framework) ??forged proof? Midterm fast file system block size consideration cylinder group internal fragmentation: sub-blocks VM context switch overhead: clock hand scan rate swap area? evolving page policies:two hands page on file system dirty code segement? address map limitation of LRU: not scanning resistent kernel fine grained protection domain copy kernel data structure for extension: kernel structure won't be left in undefined state extension doesn't do inconsistent modification of the kernel easy to recover: restart extension overhead: checking, copying kernel data structure can't recover file system inconsistency ?scan resistant sequential process communication csp http://en.wikipedia.org/wiki/Communicating_sequential_processes Totem Lamport’s causal order multiple ring: dummy message http://en.wikipedia.org/wiki/Venona_project Kerberos ??? short ticket time: short delay between password and full action RAID http://www.pcguide.com/ref/hdd/perf/raid/levels/singleLevel3-c.html http://www.pcguide.com/ref/hdd/perf/raid/concepts/perf_ReadWrite.htm small write problem raid 3? larger memory : cache will capture most read accesses? no Log based file system synchronism writes http://www.usenix.org/publications/library/proceedings/als01/full_papers/venkatesh/venkatesh_html/node21.html inode map write cost http://www.eecs.berkeley.edu/Research/Projects/CS/sprite/sprite.html Journaling vs soft updates delete directory inode first , then file inode? wafs NFS Midterm Friday: review session sequential sharing toteom: how to guruntee total order of messages. security: skip confinement problem. raid 1, raid 5, small write problem LFS paper. how LFS improve FFS Journaling and softupdates: NFS: statelessness, robustlessness, performance price lamport physical clock protocal. what's the function of NFS stale file handle: file generation number A Case For Adapting Channel Width in Wireless Networks
?? The gain is maximum in the Far-near case because sharing the same channel introduces the rate anomaly problem by which the slower flow reduces total capacity. narrower channels have a smaller per-packet relative overhead. ?? " We note that even though we do not increase total spectrum usage by splitting flows, we do increase total transmit power because narrower channels have higher power (although the same energy). It is thus an open question if the gain from such division persists in large-scale systems. " can't we ajust the transmission power accordingly? " Clearly, probing this entire search space is inefficient and we need methods that quickly converge to the optimal point. " well, it's only 32 "T (B) = B · R(B)" data rate and modulation? Distributed Robust Optimization for Communication Networks
http://en.wikipedia.org/wiki/Dual_problem Wrap testbed deployment:
ahdemo mode of madwifi driver: wlanconfig ath0 create wlandev wifi0 wlanmode ahdemo agent process deployment: deploy one complete image copy its fs online new system install business simulation games:
single player games: capitalism capitalism II industry tycoon(basicly it's a supply chain game) http://titan.ja.org/ : only one industry: holo gen! http://www.informatist.net/: text only http://www.industryplayer.com/ : real time online multilayer game. yes, real time!! non-linear and mixed integer optimization:fundamentals and applications
page 53 the gradient of equality constraints are linearly independent(i.e., x is a regular point) in general x=x(b), lambda=lambda(b) page 117 l1 or l(inf) sum of constraint violations l1-minimization problem: minimization of L1 norm: http://mathworld.wolfram.com/L1-Norm.html l inf-minimization problem: minimization of L inf norm: http://mathworld.wolfram.com/L-Infinity-Norm.html page 117: minimization is not ?? persudo convex quasi-convex subgradient: http://en.wikipedia.org/wiki/Subderivative use hessian to determine convexity(concavity) if a function: http://www.economics.utoronto.ca/osborne/MathTutorial/CVNF.HTM affine function http://mathworld.wolfram.com/AffineFunction.html page 87 ?? the master is greater than dual Characterizing the Capacity Region in Multi-Radio Multi-Channel Wireless Mesh Networks
Murali Kodialam Bell Labs, Lucent Technologies Holmdel, NJ 07733. USA muralik@lucent.com Thyaga Nandagopal Bell Labs, Lucent Technologies Holmdel, NJ 07733. USA thyaga@lucent.com a network model for multi-channel wireless network. necessary condition for rate vector static/dynamic channel assignment scheme Capacity of Multi-Channel Wireless Networks:Impact of Number of Channels and Interfaces
"In this paper, we have derived the lower and upper bounds on the capacity of static multi-channel wireless networks. We have shown that in an arbitrary network, there is a loss in network capacity when the number of interfaces per node is smaller than the number of channels. However, we have shown that surprisingly, in a random network, a single interface suffices for utilizing multiple channels, as long as the number of channels is not too large. We have also shown that interface switching delay has no impact on capacity, provided each node is provisioned with a few extra interfaces. It is part of our ongoing work to extend these results to mobile networks as well. We also plan to apply the insights gained from this work to build practical routing and MAC algorithms that approach the capacity limit " The above results are for channel model 1, in which the total bandwidth utilized doesn't vary. The network capacity is measured in terms of “bit-meters/sec” switching delay won't have much impact on the network provided that every node has a few extra interfaces. animation on linux
ktoon synfig Routing and Link-layer Protocols for Multi-Channel Multi-Interface Ad Hoc Wireless Networks
We propose a link layer protocol to manage multiple channels, and it can be implemented over existing IEEE 802.11 hardware. We also propose a routing protocol that operates over the link layer protocol, and is specifically designed for multichannel, multi-interface ad hoc wireless networks. "Several researchers have proposed MAC protocols for utilizing multiple channels [7]–[12]. These multi-channel protocols require changes to existing standards, such as IEEE 802.11, and therefore cannot be deployed by using commodity hardware. In contrast, our proposal can be implemented with standard 802.11 interfaces. Adya et al. [13] propose a link-layer solution for striping data over multiple interfaces, but their solution is designed for the scenario where number of interfaces is equal to number of channels. Bahl et al. [14] propose SSCH, a link layer solution that uses a single interface, and can run over unmodified IEEE 802.11 MAC. In contrast, our link layer protocol is designed for multi-interface networks, and we complement the link layer protocol with a routing protocol as well. " "For simplifying the description of the protocol, we assume M = 2;K = 1 for all nodes, i.e., there is one fixed, and one switchable interface" cunqing's assignment paper:
mixed interger non linear programming bender's decomposition primal/master dual problem Infeasible primal reading notes: Dichotomy: A Practical Architecture for Multi-channel IEEE 802.11 Multi-hop Networks
round robin RDT MMAC SSCH 802.11 TSF virtue of Dichotomy: flexible link level channel switching hopper/anchor seperation scatter net reading:
scatternet: optimization is not extensively studied in this area. so called optimizations are just distributed local optimization common approaches: 1. start from a plausible structure. 2. define "moves" to vary upon this structure 3. use moves to do local optimization. Bluetooth Scatternet Optimization with Hop Count Minimization: Our work builds on an optimization problem regarding the hop count minimization in scatternets. We provide a formal definition to this problem taking advantage of our matrixbased scatternet model that reflects the links and relationships between the nodes of the scatternet topology. For solving this optimization problem we defined and implemented four types of so-called moves, which are series of modifications on the master-slave relationship of the nodes. Thus, by performing a move, a node can switch to a new master (SS move), create a new piconet by becoming a master itself (SM move), give up its role of master and connect its slaves to other masters (MS move) and merge to its own piconet that of another master (MM move). Taking advantage of these moves and the matrixbased scatternet model we defined a set of algorithms capable of reconfiguring the scatternet topology such that to reduce the number of hops between the communicating nodes [5]. Distributed Self-Healing and Variable Topology Optimization Algorithms for QoS Provisioning in Scatternets In [13], loop scatternet formation is presented In [15],Tan et al. introduce the tree scatternet formation (TSF) protocol; Some other works discuss scatternet topology optimization. This issue is faced in [23] and [6] by adopting centralized approaches. "A scatternet must be formed by means of distributed algorithms running on each device. By means of these algorithms, nodes must discover their neighbors, establish links, and interconnect piconets, leading to a suitable topology." Furthermore, a scatternet formation algorithm should have the following properties: a) guarantee full connectivity to all the involved devices even in multihop scenarios, when devices are scattered in an area where not all of them are within radio range of each other (also referred to as multihop scenario); b) achieve full connectivity after limited time; c) guarantee a self-healing behavior in variable network conditions; in particular the algorithm should handle: i) entrance of new nodes in the network; ii) mobility or failure/deactivation of nodes; d) guarantee multiple and short paths between any pair of nodes; e) optimize the topology according to suitable performance metrics (e.g., overall network capacity). SHAPER-OPT: A. Step 1: Forming a Connected Tree via SHAPER B. Step 2: Optimizing the Topology via DSOA C. Step 3: Maintaining the Logical Tree for the Self-Healingness Ad hoc networking with Bluetooth: key metrics and distributed protocols for scatternet formation "a distributed algorithm for scatternet topology optimization is introduced, that supports the formation of a "locally optimal" scatternet based on a selected metric. Numerical results obtained by adopting this distributed approach to ‘‘optimize’’ the network topology are shown to be close to the global optimum." Combinatorial optimization http://en.wikipedia.org/wiki/Combinatorial_optimization metaheuristic algorithms optimization
Optimization Using Neural Networks http://portal.acm.org/citation.cfm?id=126183.126194 "The design of feedback (or recurrent) neural networks to produce good solutions to complex optimization problems is discussed. The theoretical basis for applying neural networks to optimization problems is reviewed, and a design rule that serves as a primitive for constructing a wide class of constraints is introduced. The use of the design rule is illustrated by developing a neural network for producing high-quality solutions to a probabilistic resource allocation task. The resulting neural network has been simulated on a high-performance parallel processor that has been optimized for neural network simulation." Mixed-Integer Nonlinear Programming www.gamsworld.org/minlp/siagopt.pdf " Recent research has also focused on combining of Random Search (RS), such as Tabu, Scatter Search, Simulated Annealing or Genetic Algorithms, with NLP methods. Recent implementations like OQNLP [12, 30] and LaGO [21, 22] have proven to be quite successful. Finally, the area of Disjunctive Programming uses disjunctions and logic propositions to represent the discrete decisions in the continuous and discrete space respectively. Disjunctive programs, conveniently modeled and automat- ically reformulated in big M or convex region models, give access to a rich area of applications. Widespread interest in such models has spawned a new computing environment (LogMIP [31]), developed specifically for generalized disjunctive programming. " disunctive programming www.gams.com/presentations/present_disj.pdf "" Nonlinear and mixed-integer optimization : fundamentals and applications / Christodoulos A. Floudas http://library.uh.edu/search?/Xnonlinear%20and%20Mixed%20Integer%20Optimization&CFID=298573&CFTOKEN=91041869&SORT=D/Xnonlinear%20and%20Mixed%20Integer%20Optimization&CFID=298573&CFTOKEN=91041869&SORT=D&SUBKEY=nonlinear%20and%20Mixed%20Integer%20Optimization/1%2C2%2C2%2CB/frameset&FF=Xnonlinear%20and%20Mixed%20Integer%20Optimization&CFID=298573&CFTOKEN=91041869&SORT=D&1%2C1%2C nonlinear programming, dimitri P. regular feasible: constraints satisfied feasible variation: first order feasible variation: ?? strict local minimum Weierstrass's theorem some old stuff that i don't remember: http://en.wikipedia.org/wiki/Vector_calculus http://en.wikipedia.org/wiki/Matrix_calculus something wrong with the proof in page 288: A*inv(A'*A)*A'!=I
\nabla\left|\left|x^2\right|\right| = 2x http://mathworld.wolfram.com/VectorDerivative.html Labels:
optimization Bruce Lee: Tao of Jeet Kune Do (selections from the book)
spiritdimension.com/martial-arts-texts/bruce-lee-t... Bruce Lee: Tao of Jeet Kune Do (selections from the book)Who are the super financiers of biotechnology?
http://www.nature.com/bioent/2003/030501/full/bioent733.html IP/Technology Transferhttp://www.nature.com/bioent/2008/080401/full/bioe.2008.4.html三力士(002224)首次公开发行股票招股意向书--WWW.CNLIST.COM 中国上市公司资讯网
news.cnlist.com/CnlistNewsDetail.aspx?tablename=gs... 三、简要会计报表
一）合并资产负债表主要数据 （二）合并利润表主要数据 (三）合并现金流量表主要数据 流动比率 速动比率 审计报告 古之成大
事者，规模远大与综理密微，二者阙一不可
才大器大’四字，余甚爱之。才报于器，良为知
言。
automata
computer science:
how to solve problems. course: classify problems input: alphabet. a finite set of symbols e.g.: if sigma={0, 1}, then sigma* is the set of all possible inputs. problem = language, L a subset of sigma*={inputs that has a yes answer} solving problem: a systematic step by step algorithm machine/grammar to correct answer the question for each possible input in a finite amount of time. classes of problems program machine grammar problem finite automata regular pushdown automata context free grammar program turing machine that unrestricted grammar decidable problems may not always stop procedure turing machine that unrestricted grammar undecidable problems only stops for yes answers r.e.(recursively enumerable) e.g. P1 ?= P2 |
