s*****n
发帖数: 2174 | 1 bandwidth 一般是local fitting的参数, 相当于在local多大的范围内进行regression
.
从bias-variance trade off的角度, bandwidth大, 则bias大, 但是variance小.
bandwidth小, 则bias小, variance大. Optimal bandwidth一般是通过某些预先设定的
loss function来找到的, 比如cross-validation.
从smoothing的角度, bandwidth相当于smoothing的强度, bandwidth越大, 结果越
smooth. |
|
d******i
发帖数: 7160 | 2 MSI 880G-E45 的主板。
插着一对1600的4GDDR3内存条(so 8G total)，进BIOS一看,max bandwidth是1066.
又换了一对1333的，进BIOS一看,max bandwidth还是1066。
这款主板不会连1600都不能支持吧？
倒是看见几个选项，什么"FSB/DRAM RATIO"什么的，默认被设成了auto，就是1066的
max bandwidth。手动设成1:4的话，max bandwidth就变成1600了。
如果是设置问题的话，难道这东西还不是native支持的吗？
非要手动干预一下？
请指教。谢谢！ |
|
s******t
发帖数: 71 | 3 我有一个1393x1393的稀疏矩阵，对称的，只有8308个1. 本来的bandwidth是1380，通
过行列调换之后把bandwidth减少到89. 我要求逆的矩阵是基于这个矩阵的，有一样的
structure,并且保证可逆。我发现减少bandwidth之后求逆的时间并没有加快（我只试
着用R自带的求逆函数）。 请问如果要利用减少的bandwidth，我需要什么具体的算法
去求逆然后可以加快速度呢？ |
|
g****y
发帖数: 199 | 4 ☆─────────────────────────────────────☆
seki (瓜金) 于 (Wed Mar 26 16:22:22 2008) 提到:
Is there a way to obtain this information in Matlab?
Thanks a lot
☆─────────────────────────────────────☆
szbb (深圳宝宝) 于 (Wed Mar 26 16:23:40 2008) 提到:
what is bandwidth?
☆─────────────────────────────────────☆
cockroach (@UMN) 于 (Wed Mar 26 16:33:25 2008) 提到:
[I J] = find(M);
bandwidth = max(I-J) - min(I-J) + 1
☆─────────────────────────────────────☆
seki (瓜金) 于 (Wed Mar 26 16:36:11 2008) 提到: |
|
F********E
发帖数: 1025 | 5 What does "bandwidth" mean in non-parameter fitting? Does a smaller "
bandwidth" imply a more "squeezed" shape as scale parameter mean in normal
distribution?
Any link to a easy-to-read tutorial link will also be appreciated! |
|
b****l
发帖数: 9 | 6 小弟恳请大牛推荐几款bandwidth的Op-Amp吧
unity-gain bandwidth大于500MHz或者1GHz
多谢了！ |
|
F********E
发帖数: 1025 | 7 What does "bandwidth" mean in non-parameter fitting? Does a smaller "
bandwidth" imply a more "squeezed" shape as scale parameter mean in normal
distribution?
Any link to a easy-to-read tutorial link will also be appreciated! |
|
B****n
发帖数: 11290 | 8 For a non-parametric fitting problem, the method usually involves smoothing
the data. The bandwidth determines the degree of smoothness; the larger the
bandwidth, the more smooth the fitting curve. |
|
s*******g
发帖数: 483 | 9 in nonparametric density estimation, bandwidth refers to the width of the
kernel (such as the gaussian) that is used to build the estimate. The larger
the bandwidth, the coarser the estimate which may corresponds to larger-
scale behaviour of the distribution |
|
B****n
发帖数: 11290 | 10 你應該先問普通kernel density estimator的bandwidth和n的關係
然後bootstrap 的sample需要比較大一點的bandwidth 因為standard bootstrap
是取完放回再抽 因此要smooth的時候 需要比原來的sample大的window |
|
F********E
发帖数: 1025 | 11 What does "bandwidth" mean in non-parameter fitting? Does a smaller "
bandwidth" imply a more "squeezed" shape as scale parameter mean in normal
distribution?
Any link to a easy-to-read tutorial link will also be appreciated! |
|
z*******n
发帖数: 1034 | 12 On the surface, it's not a lot different than the company's current LTE: the
icon is the same, the phones are basically the same (we'll actually get to
a list a little bit later), and no modifications are needed to current plans
. XLTE is about making sure customers in big cities – places that are
generally congested – always have the bandwidth and speed they need to get
things done by allowing access to both 700 MHz and AWS spectrums at the same
time. This, in theory, should also alleviate some... 阅读全帖 |
|
t*****a
发帖数: 5180 | 13 所有的bandwidth都用在舔米国，求做儿皇帝上了。 |
|
d********r
发帖数: 199 | 14 大家想开个人网站的互相帮助一下如何？我可以提供server+bandwidth
各位大侠，首先声明，本人可是网站新手，有个提议，请高手赐教。
我正在计划开一个个人网站。网站是从0做起，本来想租个空间的，
后来想想，不如先在家里run一段时间，等流量上来后再租好了。
我的计划是这样的，先把server 在自己家里run,
家里的带宽是comcast的cable service, download/upload=6Mbps/384Kbps
只要没有大的图片的话，一般性的访问速度还是够快的。
当然，IP地址是动态的。不过，实际情况下，我发现comcast显然对客户端的
MAC IP地址做了cache，如果我掉线或断电重连cable modem,只要这边不换
设备的话，地址始终不变，有一次我断掉modem达半天左右，再接上发现IP地址
仍然没变，所以，这个动态IP地址基本上可以当成准静态的使用。
即使IP地址变了话，我的域名用的动态IP service,经我测试，大概可以在2-3个小时
内得到更新，也就是说，万一有IP地址变化，也就是损失2-3个小时的访问时间吧。
不过，我觉得这个情况发生的概 |
|
c**t
发帖数: 2744 | 15 I'm using a linux device to stream video over Wi-Fi. The problem I had is
that video quality isn't good: delay, choppy. Ran "bsm-ng", the live
bandwidth monitor, I found that the Rx & Tx of wlan0 rapidly changed from
500KB/s to 0KB/s. How to troubleshoot the issue? |
|
c**t
发帖数: 2744 | 16 If this is expected behavior, any suggestion to improve the video quality
with 2.4G? I have very good bandwidth and signal (as follow), expecting
much better video output. Thanks for any hint!
I do noticed that my device can "see" about 25 APs in range, and 11 of them
are on the same channel to the one I'm connecting to. Is the overlap the
root cause?
wlan0 IEEE 802.11abgn ESSID:"XXXNet"
Mode:Managed Frequency:2.412 GHz Access Point: C0:25:5C:55:C9:29
Bit Rate=72.2 M... 阅读全帖 |
|
p**********y
发帖数: 13 | 17
look for current feedback amps. they don't suffer the gain vs. bandwidth
tradeoff facing voltage feedback amps and many of them are good for very
high frequencies.
just be careful with layout, though. vfb opamps oscillate like crazy. |
|
h***i
发帖数: 3844 | 18 比如standard bootstrap 是 bandwidth 为 n 的 －（１／２）方
直觉上怎么理解？ |
|
h***i
发帖数: 3844 | 19 那是不是说
standard bootstrap 和bootstrap from kernel density estimator with bandwidth
=n^{-1/2} 是一码事？
多谢，想破头了，实在是愚钝
这是for quantile variance
我翻了些paper, 比如 Angelis 和 Young的文章，没理出头绪 |
|
h***i
发帖数: 3844 | 20 翻翻fan jianqing 的lcal polynomial 就可以了
bandwidth是调节goodness of fit 和overfitting的。 |
|
F********E
发帖数: 1025 | 21 Didn't get the book yet. Just wondering, does the value of the automatically
optimized bandwidth (by Matlab) imply something related to the shape? |
|
w*******y
发帖数: 60932 | 22 Link:
http://www.maxcdn.com/
MaxCDN 1 cent Coupon code- testdrivemaxcdn will give you an account, plus
1TB of bandwidth for only 1 Cent. You have up to 12 months to use it.
Use your coupon code within the next 7 days on MaxCDN or HDDN. (Coupon
Expires: 3-13-12)
quote
contact info - 1-877-MAXCDN1 MaxCDN experts are on call to assist
you Mo-Fr 9am-7pm (PST).
s***[email protected]
|
|
d****n
发帖数: 12461 | 23 ☆─────────────────────────────────────☆
atnalta (职业debugger) 于 (Fri May 4 14:45:08 2012, 美东) 提到:
一个朋友买了垃圾油，让我帮着刷Firmware。趁这个机会昨天晚上做了一些测试。
我手头没有300M的2.4G卡。今天晚上回家后用公司笔记本在做WIFI传输速度的测试。
手头有四个路由器，WRT-54GS, E2000, RT-N56U和垃圾油。WRT-54GS就不比较了。
E2000和垃圾油都刷成dd-wrt，同一个版本。
都采用Channel 1, N only, 20Mhz。同样的输出功率74mw
RT-N56U是stock firmware
信号强度
RT-N56U 平均51db
E2000平均62db
垃圾油平均75db
有线传输速度
RT-N56U 980Mbps
E2000 760Mbps
垃圾油 560Mbps
150M 2.4G, 20MHz
RT-N56U 56Mbps
E2000 42Mbps
垃圾油 26Mbps
无线测试用的是MBA。用jperf和Linux... 阅读全帖 |
|
k*********g
发帖数: 791 | 24 有限元是个烂方法的原因：有限体积法比有限元快10倍左右。
数值计算分2个阶段，“矩阵形成”和“矩阵迭代”；矩阵形成消耗的时间大概是25次
迭代（这种估算早该在教科书中出现，却在过去几十年内没人说这样的事），而通常的
一个time dependent问题的每个time marching内，需要几百次的迭代（假如时间步长
够小的话，否则几万次迭代也可能收敛不了）；所以，一个算法的效率，关键看它的矩
阵迭代，which关键看它的bandwidth。
譬如，同样一个3维问题，有限元形成的矩阵的bandwidth有1、2百，而有限体积法的
bandwidth只有1、2十；我们知道，
矩阵迭代的过程，就是矩阵对角化过程，初始的时候bandwidth大的矩阵，需要走过一
个更漫长的路变成bandwidth=1
的矩阵。
当然有限体积法，虽然能轻松beat有限元，但它毕竟也是一个很老的数学智商很低的一
个方法（有限体积法1965年诞生，
有限元1953年诞生），真正强大的现代的数值方法是。。。。。。
谱元法 spectral element method |
|
a*****a
发帖数: 1429 | 25 一个朋友买了垃圾油，让我帮着刷Firmware。趁这个机会昨天晚上做了一些测试。
我手头没有300M的2.4G卡。今天晚上回家后用公司笔记本在做WIFI传输速度的测试。
手头有四个路由器，WRT-54GS, E2000, RT-N56U和垃圾油。WRT-54GS就不比较了。
E2000和垃圾油都刷成dd-wrt，同一个版本。
都采用Channel 1, N only, 20Mhz。同样的输出功率74mw
RT-N56U是stock firmware
信号强度
RT-N56U 平均51db
E2000平均62db
垃圾油平均75db
有线传输速度
RT-N56U 980Mbps
E2000 760Mbps
垃圾油 560Mbps
150M 2.4G, 20MHz
RT-N56U 56Mbps
E2000 42Mbps
垃圾油 26Mbps
无线测试用的是MBA。用jperf和Linux服务器iperf连接。和路由器楼上楼下隔两个房间
，其中一堵墙是承重墙。垃圾油的速度太悲催了。测试的最大速度都没有超过30Mbps，
不到RT-N56U的一半。
测完40Mhz 300M模式后，会建议朋友出... 阅读全帖 |
|
f*********t
发帖数: 37 | 26 I agree with you at this point. Symbol rate is not signal bandwidth. To
reconstruct signal, we need sampling rate at least twice of the signal
bandwidth. In real system, using just twice of the signal bandwidth sampling
rate may casue a lot of design issues and sampling rate is usually higher
than this.
Bandwidth here means single side baseband bandwidth.
information
but
and
( |
|
e*u
发帖数: 10016 | 27 美军为日本救灾屏蔽各大主流网站！——根据CNN得到的一封电子邮件，美军已经屏蔽
了一些主流网站，表面理由是“为日本救灾让出带宽”。该消息已经得到美军战略指挥
部的证实。
Washington (CNN) -- The U.S. military has blocked access to a range of
popular commercial websites in order to free up bandwidth for use in Japan
recovery efforts, according to an e-mail obtained by CNN and confirmed by a
spokesman for U.S. Strategic Command.
The sites -- including YouTube, ESPN, Amazon, eBay and MTV -- were chosen
not because of the content but because their popularity among users of
military c... 阅读全帖 |
|
s******r
发帖数: 5309 | 28 Evolution of Internet in China
2001-01-01
On September 20, 1987, Professor Qian Tianbai sent China's first E-mail
titled "Crossing the Great Wall to Join the World," marking the beginning of
the use of Internet by Chinese. Professor Qian was in charge of the
Internet project of CANET (Chinese Academic Network), a scientific research
project launched by the Beijing Municipal Computer Application Research
Institute in 1986 in cooperation with Karlsruhe University of former Western
Germany. The mai... 阅读全帖 |
|
k**o
发帖数: 15334 | 29 微软是被迫买akamai公司的bandwidth分流，因为微软本身不是做bandwidth的，
当时没有那么多bandwidth来吸纳DDoS。碰巧akamai公司用的是linux而已。
在这个事件里面linux没有任何特殊性，只是serve up了一些static page。
windows server一样可以，只是微软当时没有防备DDoS，没有那么多硬件
和bandwidth，估计有了这次教训之后，现在没人能DDoS微软了。
除非你的意思是linux能不被DDoS？ |
|
j**z
发帖数: 109 | 30 Good analysis from a typical MBA outlook, who might never invest companies
like google, netflix, amazon in early stage because all the bad side you
mentioned. For investors, who have to understand the nature of the business,
the potential and the execution to make an educated bet.
The key of GOGO's success is really depending on how well it runs in the
next 12 month, after launching the new 2Ku. To me, GOGO and VSAT/Panasonic
are very different companies. VSAT/Panasonic makes living by building ... 阅读全帖 |
|
j**z
发帖数: 109 | 31 Good analysis from a typical MBA outlook, who might never invest companies
like google, netflix, amazon in early stage because all the bad side you
mentioned. For investors, who have to understand the nature of the business,
the potential and the execution to make an educated bet.
The key of GOGO's success is really depending on how well it runs in the
next 12 month, after launching the new 2Ku. To me, GOGO and VSAT/Panasonic
are very different companies. VSAT/Panasonic makes living by building ... 阅读全帖 |
|
j**z
发帖数: 109 | 32 Good analysis from a typical MBA outlook, who might never invest companies
like google, netflix, amazon in early stage because all the bad side you
mentioned. For investors, who have to understand the nature of the business,
the potential and the execution to make an educated bet.
The key of GOGO's success is really depending on how well it runs in the
next 12 month, after launching the new 2Ku. To me, GOGO and VSAT/Panasonic
are very different companies. VSAT/Panasonic makes living by building ... 阅读全帖 |
|
j**z
发帖数: 109 | 33 Good analysis from a typical MBA outlook, who might never invest companies
like google, netflix, amazon in early stage because all the bad side you
mentioned. For investors, who have to understand the nature of the business,
the potential and the execution to make an educated bet.
The key of GOGO's success is really depending on how well it runs in the
next 12 month, after launching the new 2Ku. To me, GOGO and VSAT/Panasonic
are very different companies. VSAT/Panasonic makes living by building ... 阅读全帖 |
|
|
h**2
发帖数: 2841 | 35 http://en.wikipedia.org/wiki/Hdmi
HDMI version 1.0–1.2a 1.3 1.4
Maximum signal bandwidth (MHz) 165 340 340[50]
Maximum TMDS bandwidth (Gbit/s) 4.95 10.2 10.2
Maximum video bandwidth (Gbit/s) 3.96 8.16 8.16
Maximum audio bandwidth (Mbit/s) 36.86 36.86 36.86
Maximum color depth (bit/px) 24 48[A] 48
Maximum resolution over single link at 24-bit/px[B] 1920×1200p60 2560×
1600p75 4096×2160p24
Maximum resol |
|
|
|
i****s
发帖数: 375 | 38 我在QUORA 上看的一个问答，这个东西应该和我想做的类似
How much would it cost to develop a video sharing website like YouTube in
India?
有个哥们回答：
The costs are extremely high for a viable ad-supported free model. Since
video sharing sites are very bandwidth and server usage intensive, so unless
you already own company that can lay down its own fibre optic cables..
costs will remain high.
Bandwidth is comparably very expensive in India than North America, Europe,
South Korea or Singapore. Even if you manage to get bandwidt... 阅读全帖 |
|
c*u
发帖数: 916 | 39 Of course dispersion will affect a 30 nm bandwidth pulse in 4 m of fiber,
especially if the center wavelength is not 1310 nm or it's specified
zero-dispersion wavelength. 30 nm is very wide bandwidth for most
applications. Let's say dispersion 5 ps/nm/km, then for 4 m and 30 nm, it
is almost 0.6 ps. The bandwidth is narrowed to almost 6 nm.
On the other hand, it's very rare that a fiber can't support 30 nm
bandwidth. |
|
f*********t
发帖数: 37 | 40 I am not 高人.
You cannot know the modulated signal bandwidth just from the symbol rate.
This really depends on your modulation scheme. Like in GSM system, symbol
rate is 270.08333kHZ. The baseband bandwidth of the signal is only 100kHz.
When your symbol rate is higher than two times of the bandwidth of a
baseband signal, then ISI cannot be avoided. We need equalization to detect
signal.
note: GSM channel is usually known to be 200KHz, this number is passband
bandwidth after modulated on radio fre |
|
f*********t
发帖数: 37 | 41 look at the notes in my previous post, I said 100kHZ is the baseband
bandwidth. After up convert to carrier frequency, then it becomes 200kHz. I
believe in Nyquist theorem, the bandwidth is baseband bandwidth.
"as for nyquist, if the bandwidth is W Hz, then you can reconstruct the
waveform with >2W/s sample rate, right"
If this is how you get your conclusion, then we are not talking about the
same thing. In digital communication, we don't really reconstruct the whole
tranmitted signal. All we do |
|
w*******y
发帖数: 60932 | 42 I've been using NearlyFreeSpeech.NET:
https://www.nearlyfreespeech.net/
for over a year, and I can safely say that this is without doubt the most
affordable and reliable option for hosting small websites.
Even though NFSN don't advertise it this way, their NearlyFreeSpeech.NET:
https://www.nearlyfreespeech.net/
is very similar to that of a cloud hosting service. The only difference is
that NFSN only charges pennies for what most cloud hosting services charge
dollars for. As of January 2011, ... 阅读全帖 |
|
w*******y
发帖数: 60932 | 43 TheCubeNet has released its ad for Black Friday Specials:
Here's the link:
Link:
http://www.thecubenet.com/blackfriday.htm
Deal is only live on 11/24 @ 9pm PST
Black Friday Doorbuster specials*
Unlimited Monthly (unlimited bandwidth, unlimited speed, 1118+ retention, US
& EU Servers, 24/7 Support)
$4.99/month 20 SSL encrypted connections (500 available)
Block Plans (no expiration date, 50 SSL encrypted connections, 1118+
retention, US & EU Servers, 24/7 Support)
$1.79/month 25GB unlimited speed ... 阅读全帖 |
|
F***Q
发帖数: 6599 | 44
your home uplink speed = ISP bandwidth/total (potential) users
ISP bandwidth depends on the number of physical connections, the data
bandwidth of each connection, and encoding schemes. all of these are
variables in the equation.
the latest record for optical-fiber-based communication is 1.05 Petabit/s [
1]. what the ISPs can provide has not yet been close to the physical limit,
and the limit itself is a moving target.
[1]
http://en.wikipedia.org/wiki/Fiber-optic_communication#Record_s |
|
g********g
发帖数: 2172 | 45 You really should make a 2-5 Kb logo. It won't cause bandwidth problem for
you as long as you do not used image in large scale. At your current traffic
, you won't use up the bandwidth assigned to you anyway, right? Then, why
not use it. If a small logo image will cause bandwidth problem for you, you
are successed already. |
|
t******t
发帖数: 121 | 46 Baidu Announces Fourth Quarter and Fiscal Year 2010 Results 01/31 04:
30 PM
BEIJING, Jan. 31, 2011 /PRNewswire-Asia/ -- Baidu, Inc. (BIDU:$108.63,00$2.
09,001.96%) , the leading Chinese language Internet search provider, today
announced its unaudited financial results for the fourth quarter and fiscal
year ended December 31, 2010(1).
(Logo: http://photos.prnewswire.com/prnh/20081103/BAIDULOGO)
Fourth Quarter and Fiscal Year 2010 Highlights
Total revenues in the fourth quarter of 2010 were ... 阅读全帖 |
|
s*****n
发帖数: 5488 | 47 Published: Monday, 8 Oct 2012 | 4:27 PM ET Text Size
By: John Melloy
Executive Producer, Fast Money & Halftime
Twitter
14
LinkedIn
0
Share
A single mysterious computer program that placed orders — and then
subsequently canceled them — made up 4 percent of all quote traffic in the
U.S. stock market last week, according to the top tracker of high-frequency
trading activity. The motive of the algorithm is still unclear.
The program placed orders in 25-millisecond bursts involving about 500
stocks, ... 阅读全帖 |
|
