l**b
发帖数: 457 | 1 这个我可以完全反给你啊，Hash都比Array还省空间的话，要Array干嘛？这个要看情况
的吧。你怎么知道一定是sparse的array。
我不知道你们有没有实际的工作经验，你们自己去看看自己写的code，那个array可能
是sparse的。我还真没见过那个initializ一个100个element的array然后只用一个。
我不是反驳我多厉害，我基础是很差，我不否认，写这个是分享经验。不是吐槽面试官。 |
|
q******8
发帖数: 848 | 2 不知道lz面的时候问没问initinityarray的定义。我觉得如果那人说请现实一个
inifinityArray，这个明显没法写吧，如果定义好了，需求明确，那估计他就会把
sparse array说出来了。我觉得前面有人说人家都说long index了，一定是sparse
array，我觉得这也太。。。。你怎么知道啊。
我觉得如果光看lz提供的面经，问题应该是lz没问仔细需求，一下就开始coding了
总之感谢lz宝贵面经 |
|
x*n
发帖数: 33 | 3 sparse matrices 咋表示 compressed sparse row or column ? 还是自己随便定义啊 |
|
g*******n
发帖数: 2198 | 4 【 以下文字转载自 Faculty 讨论区 】
发信人: goldenman (戈登曼), 信区: Faculty
标 题: 借版问个least square的问题
发信站: BBS 未名空间站 (Wed May 18 18:50:25 2016, 美东)
http://math.stackexchange.com/questions/1664018/is-there-an-ite
主要两个问题
1.U比较大的时候 没法直接算的时候怎么处理 LSMR只能解决Sparse的时候 如果不是
Sparse 是不是没办法了
2.后半个问题 如何iteratively得到theta， LSMR虽然简化了算法 但还是要存整个U
如果U过大 有memory问题的话怎么处理 我看下面的回答 用gradient descent根据
error function来算，但error function还是需要以前的data y_i u_i 有没有办法在
新的data进来的时候只用现有的data来update theta同时minimize error of whole
history, 比如theta_(i... 阅读全帖 |
|
发帖数: 1 | 5 https://www.linkedin.com/in/krishna-pai-987571a
就是这位哥们,上来先mumble 一下自己的project,各种machine learning big data,然
后就开始冷笑呵呵
第一题:
write the interface of cache
我猜他是在问LRU cache,于是我说用python,给出get(), put(),因为我不知道他是不是
用key value store,所以就没有没有给出function signature,然后他问能不能java或
者c++,因为他不熟悉python,然后说了一些abstract class的东西,我说interview我一
般用python,三哥然后hehe,电话我这边听到他笑得好开心
第二题:
use map-reduce to implement matrix multiplication for large and sparse
matrix C=A*B
这个上来就刷新了我对大公司面试的理解,我以为都是考leetcode或者system design
我面的不是senior... 阅读全帖 |
|
n**a
发帖数: 41 | 6 MERL is seeking well qualified applicants for an internship in the Digital
Communications group. This group is embarking on a new research project to
study the application of recent advances in sparse signal processing to
wireless/wired communication systems. Ideal candidates will devote all his/
her attention to this research project.
Responsibilities:
• Initiate and conduct independent research in the application of
sparse signal processing to communications problems, such as wireless
c |
|
l****r
发帖数: 5317 | 7 umm, seems the translator felt hard in this part too
月光是隔了树照过来的，高处丛生的灌木，落下参差的斑驳的黑影，却又像是画
在荷叶上。
As the moon shone from behind them, the dense trees on the hills threw
checkered shadows, dark forms loomed like devils, and the sparse, graceful
shadows of willows seemed painted on the lotus leaves.
every
the
our
him.
track
of
azure
And
carried
in
to
sparse,
moonlight
rhythm |
|
f*******y
发帖数: 423 | 8 nonparametric方面很多东西可以搞得，smoothing这个是很重要也是比较基本的一块，
还有nonparemetric的时候往往要研究asymptotical的情况。不过high dimension的时
候好像是麻烦点，最近听了个seminar，他们的结论只能在sparse的情况下才能用，不
是sparse的情况下结论不成立。还听过个random matrix theory的，这个基本不懂，就
只知道他matrix，eigenvalue搞来搞去，但不知道在干什么，哈哈 |
|
f*******y
发帖数: 423 | 9 nonparametric方面很多东西可以搞得，smoothing这个是很重要也是比较基本的一块，
还有nonparemetric的时候往往要研究asymptotical的情况。不过high dimension的时
候好像是麻烦点，最近听了个seminar，他们的结论只能在sparse的情况下才能用，不
是sparse的情况下结论不成立。还听过个random matrix theory的，这个基本不懂，就
只知道他matrix，eigenvalue搞来搞去，但不知道在干什么，哈哈 |
|
m****s
发帖数: 402 | 10 谁能给详细讲解一下lz这几个题和概念?
1) 模式识别的Linear Discriminant, Neural Network, SVM, Kernel, unsupervised learning
2)求2叉树深度
3)基于mutual information有关的feature selection
4)为什么Least Square基于高斯分布是最优的
5)sparse matrix的sparse vector的排序算法
6)3百万维的用户数据, HMM, Bayesian, SVM, ANN各有什么算法优势
【 以下文字转载自 JobHunting 讨论区 】
发信人: diamondfan (diamond), 信区: JobHunting
标 题: Yahoo!实验室Data Mining and Research Group Offer
发信站: BBS 未名空间站 (Tue Feb 13 16:05:11 2007)
刚刚收到HR的口头Offer,达到心里价位，当场答应了。找工作真的很累，所以不想折腾
了，毕竟这个职位我很满意。
2周前on site，6个人轮番轰炸。9 |
|
S******o
发帖数: 111 | 11 ☆─────────────────────────────────────☆
mytbbs (廉颇老矣) 于 (Wed Feb 14 14:33:39 2007) 提到:
谁能给详细讲解一下lz这几个题和概念?
1) 模式识别的Linear Discriminant, Neural Network, SVM, Kernel, unsupervised learning
2)求2叉树深度
3)基于mutual information有关的feature selection
4)为什么Least Square基于高斯分布是最优的
5)sparse matrix的sparse vector的排序算法
6)3百万维的用户数据, HMM, Bayesian, SVM, ANN各有什么算法优势
发信人: diamondfan (diamond), 信区: JobHunting
标 题: Yahoo!实验室Data Mining and Research Group Offer
发信站: BBS 未名空间站 (Tue Feb 13 16:05:11 2007)
刚刚收到HR的口头Off |
|
B****x
发帖数: 17 | 12 http://blog.sina.com.cn/s/blog_661c16d00100kc51.html
机器学习大家谈（转贴）(2010-03-17 21:42:19)转载
标签：杂谈 分类：学术科研
闲着无事，想写点一些我所了解的machine learning大家。由于学识浅薄，见识有限，
并且仅局
限于某些领域，一些在NLP及最近很热的生物信息领域活跃的学者我就浅陋无知，所以
不对的地方大
家仅当一笑。
Machine Learning 大家(1)：M. I. Jordan
在我的眼里，M Jordan无疑是武林中的泰山北斗。他师出MIT，现在在berkeley坐
镇一方，
在附近的两所名校（加stanford）中都可以说无出其右者，stanford的Daphne Koller
虽然也
声名遐迩，但是和Jordan比还是有一段距离。
Jordan身兼stat和cs两个系的教授，从他身上可以看出Stat和ML的融合。
Jordan 最先专注于mixtures of experts，并迅速奠定了自己的地位，我们哈尔滨
工业大
学的校友徐雷跟他做博后期间，也在这个方向上沾光不少... 阅读全帖 |
|
x*********n
发帖数: 28013 | 13 想来想去没想明白。
2个interface是这样的，用的是public IP
interface FastEthernet0/0
description **** MPLS data interface *****
ip address 10.99.32.1 255.255.255.128
ip pim sparse-dense-mode
duplex full
speed 100
!
interface FastEthernet0/1
description **** Voice interface *****
ip address 10.99.33.1 255.255.255.128
ip pim sparse-dense-mode
duplex full
speed 100
WAN IP，multilink出去。
interface Multilink1
ip address 65.112.128.118 255.255.255.252
ppp multilink
ppp multilink fragment disable
ppp multilink group 1
然后这2个... 阅读全帖 |
|
G*****7
发帖数: 1759 | 14 function fast_grid_lookup
%% prepare the data
% randomly generate the query points
num_points = 20e6;
points = rand(num_points,2);
% specify the evenly-spaced mesh grid
num_divs = 1028; % per side
num_cells = num_divs^2;
grid_spacing = 1/(num_divs-1);
% you do not have to instantiate the grid points by
% grid = meshgrid(linspace(0, 1, num_div), linspace(0, 1, num_div));
%% find the enclosing cell of each point
tic;
points_in_cell = ceil(points/grid_spacing)+1; % damn you, 1-based matlab
toc;
%% ... 阅读全帖 |
|
w***g
发帖数: 5958 | 15 sparse blas. CPU做因为cache的问题讨不了太多好。
或许可以通过数学方式把几层sparse合成一层优化。
按风水轮流转的原理，这方面将来肯定要突破的。
不然没法做symbolic inference。 |
|
b*****i
发帖数: 58 | 16 if N<>M, then x is given as the Least Square Solution. For sparse matrix, you
can use the sparse mode in Matlab for it. But no idea about its parallel
version. |
|
s*******g
发帖数: 4 | 17 blitz++ is designed to be a multi-dimensional array library (a very fast one),
not
a linear algebra library (such as boost::ublas or lapack++). For instance,
blitz++
does not provide sparse matrices. A look at lapack++ website
http://math.nist.gov/lapack++/
seems to imply that it is not actively developed (last update in 2000).
boost::multi_array also provides multi-dimensional arrays, but it is much
slower
than blitz++. I have only used boost::ublas for linear algebra (e.g. sparse
matrices)
a |
|
w*******e
发帖数: 6 | 18 我觉得如果对于sparse matrix 不是很熟悉的话, 有个 叫作sparse 的C程序库很好用,
它也提供了Fortran 的界面. 如果想在自己的程序中用的话, 这个库还是比较容易
上手的. 比较容易找到的是1.3 (netlib上就是这个版本), 大概因为是比较老的C程
, 所以编译常常会有些问题. 幸运的是, 最近作者推出了新版本, 所以编译也不是问
题了. 如果感兴的话, google一下就可以找到了. |
|
l*****i
发帖数: 3929 | 19 More information needed: Is the matrix sparse? Is it positive definite?...
If symmetric, you can try ma27 or lu factorization ("\"). If it's sparse and
positive definite, try CG.
'.
free |
|
c*******h
发帖数: 1096 | 20 what you are in front of is an entire area of sparse matrix computation.
unless you have taken a course on it, it is unlikely that you understand
the details from just a few words from me.
in short, what i was proposing to you are a few iterative algorithms to
solve your sparse linear system. you can use the `help' command in matlab
to read the instructions on how to use these algorithms, to get a feel
which algorithm is suitable for your matrix, and to get to know (and
try) more algorithms from |
|
b***e
发帖数: 38 | 21 Even for a 1000 by 1000 dense matrix, it takes a LU solver written in C less
than 5 second on a 2.6 GHz processor. Besides, the LU solver is not
optimized at all. It seems your machine needs to be upgraded. The
performance of sparse LU is problem dependent. It's hard to predict the CPU
time just by the number of unknowns. Again on a 2.6 GHz CPU, you can solve a
matrix with 100,000 unknowns in less than 10 seconds or even 1 second with
a sparse LU solver written in C if the problem is nice.
Ma |
|
r*****y
发帖数: 1 | 22 Directed Transmission Method (DTM),
a fully asynchronous approach to solve sparse linear systems in parallel
DTM is a fully asynchronous, scalable and continuous-time iterative
algorithm to solve the arbitrarily-large sparse linear system whose
coefficient matrix is symmetric-positive-definite (SPD). DTM is able to be
freely running on the heterogeneous parallel computer with arbitrary number
of processors, which might be manycore microprocessors, clusters, grids,
clouds, and the Internet.
prepr |
|
b*********h
发帖数: 46 | 23 希望可以在这里找到提示。我有一个大型非对称sparse matrix，
可以计算其最大的若干个eigen value／vector，希望用这些结果
构造一个low rank的sparse pseudoinverse，想知道的是，这样
构造出来的和用svd构造出来的pseudoinverse相比有什么问题没
有？主要不想计算svd，感觉应该比eigen value decomposition
难算一些。当然还需要对eigen vector做QR分解，不是很清楚究
竟是否更efficient些，多谢达人指点。 |
|
S****8
发帖数: 401 | 24 现在一个项目用的一个 C++ template library叫FLENS 类似于eigen，
我需要储存很多个大小相同的矩阵，在后面的计算中备用，
想法很简单，用 std::vector 容器，
std::vector Dmatrices;
std::vector Smatrices;
// some operations, e.g.
Dmatrices.push_back(Dmatrix1);
Smatrices.push_back(Smatrix1);
结果，dense matrix这样搞是可以的，但是sparse matrix 这样链接的时候就有问题，
我在想是不是因为 std::vector 要求每个element是一样的，而sparse matrix尽管
dimension一样，但本质上而是大小不同的？ |
|
s**p
发帖数: 207 | 25 ok,
i have a symmetric matrix A and I want to solve Ax=b
if I can get its upper left matrix, let's say B, (the diagonal term is 1/2
Aii)
then B+B'=A; if I have a sparse transformation matrix T: B'=TB
Ax=b=> (B+B')x=b => (I+T)Bx=b
Bx=(I+T)^-1 b
(I+T)^-1 is easy. because T is sparse
so x is obtained in O(N^2), inversion free.
i guess T must be dense, otherwise, too simple to be true
bet
e
transfo
nothing |
|
c****n
发帖数: 2031 | 26 粗略的说
L2刻画的函数：bumpy and broad
L1刻画的函数：peaky and concentrated
L2函数的数值的分布是Gaussian（意思是非零数值dense）
L1函数的数值的分布是Laplacian（非零数值sparse）
所以无所谓优缺点。但目前流行的观点是L1比L2好，是因为从哲学角度讲，
information
总是在某种合适的变换下是sparse的，所以用L1 model比较合适 |
|
m****h
发帖数: 13 | 27 if your solution/matrix do not have sparsity, there is no meaning to use
wavelet approach for numerically solve PDE.
Advantages for Wavelet approach:
1. Wavelet method combine(O(N)) the advantages of both spectral (O(N Log_2 N
)) and finite difference methods
and allow both space and time dependent coefficients;
2. Large classes of operators and functions are sparse or sparse to high
accuracy when transformed into the wavelet domain;
3. It is suitable for problems with the multiple spatial |
|
k****e
发帖数: 647 | 28 A more interesting problem can be
if the matrix is sparse, lots of 0, and the matrix is in the form of sparse
representation (3 long columns), how to find an efficient solution. |
|
d******e
发帖数: 7844 | 29 是Sparse Matrix么？
是的话用Sparse Matrix表示 |
|
r***y
发帖数: 25 | 30 Their macro architecture, instead of their micro ones, are more interesting
. From a glimpse of the diagram (it is really a glimpse in all of their
videos, LOL), they have single units from all brain regions (visual cortex,
thalamus, BG, etc) involved in a task and establish hardware connections
between these units. The single unit could be any currently successful
neuron model (Though in personal opinion, the Izhikevich version is best
suited in this large-scale network approach). The connecti... 阅读全帖 |
|
r***y
发帖数: 25 | 31 Their macro architecture, instead of their micro ones, are more interesting
. From a glimpse of the diagram (it is really a glimpse in all of their
videos, LOL), they have single units from all brain regions (visual cortex,
thalamus, BG, etc) involved in a task and establish hardware connections
between these units. The single unit could be any currently successful
neuron model (Though in personal opinion, the Izhikevich version is best
suited in this large-scale network approach). The connecti... 阅读全帖 |
|
u*****o
发帖数: 1224 | 32 熊！抓住！
我也觉得有点问题，感觉完全discard一组数据不太make sense，我胡乱算了半天也没
有什么conclusive的结果。。。
什么是sparse indicator？这两个string对每个用户都有记录，这也算sparse吗？ |
|
s****h
发帖数: 3979 | 33 新手请教一个scala+spark问题:
数据格式: userid itemid1,itemid2,itemid3,...
sample:
5 7,13,20,27
12 7,10,27,33
需要做的是,每两个userid的item vector算cosine, 每个user, sort其他user by
cosine value.
其实就是一个KNN.
我想得是
a. 一行行读进来,每行一个sparse vector
or
b. 一起读进一个sparse matrix
两种方式都不会.
多谢多谢。
scala+spark经验只限于根据spark summit上的alswr code照猫画虎，处理了一下自己
的数据。 |
|
H****E
发帖数: 254 | 34 https://databricks.com/blog/2014/07/16/new-features-in-mllib-in-spark-1-0.
html
"So storage-wise, the sparse format is better than the dense format when
more than 1/3 of the elements are zero. However, assuming that the data can
be fit into memory in both formats, we usually need sparser data to observe
a speedup, because the sparse format is not as efficient as the dense format
in computation. Our experience suggests a sparsity of around 10%, while the
exact switching point for the running time... 阅读全帖 |
|
c***u
发帖数: 4107 | 35 刚刚学到, One-hot encoding, 转换成sparse vector, 然后再把sparse vector给hash
了 |
|
g****u
发帖数: 25 | 36 骑驴找马今天面了一个感觉牛逼哄哄的startup, 已挂，满满的挫败感。最大的感受是
平时做项目的时候machine learning的算法都是抓来用用，那个好用用哪个，但对算法
的理解都很肤浅，没有深入思考过。大牛们看看下面几个问题怎么回答才好？
1）第一个问题是我有一个项目用mape来evaluate模型，面试官问我为什么不能用rmse
，我说mape是客户要求的，所以我就没多想，我实在不好意思讲我当时瞎扯了点啥，
太打脸了
2）面试官问我favorite的算法是什么，我说没有，平时logistic regression和random
forest用的比较多，然后面试官就问我什么时候用lg什么时候用rf。虽然我理论上知
道一点各自的特点比如lg对outlier比较敏感，模型interpretability比较高, rf是non
-parametric的所以对outlier不敏感，不用担心colinearity的问题之类的，但用到具
体问题的时候，我就说不上来了，因为其实我是抓来都用用，如果两个算法
performance差的比较大，我就直接用好的那个。我临时就凭感觉说如果featu... 阅读全帖 |
|
i***s
发帖数: 39120 | 37 It is a finding certain to be hotly disputed – at least among half the population. According to the latest research, women are brighter than men.
For the first time in IQ testing, psychologists have found that female scores have risen above those of men.
Since IQ testing began a century ago, women have been as much as five points behind, leading psychologists to suggest embedded genetic differences.
But that gap has been narrowing in recent years and this year women have moved ahead, according t... 阅读全帖 |
|
y****i
发帖数: 1050 | 38 The T-34 came to symbolise the effectiveness of the Soviet counterattack
against the Germans. One of the first known encounters with a T-34 involved
the 17th Panzer Division near the Dniepr River. The T-34 crushed a 37 mm
anti-tank gun, destroyed two Panzer IIs, and left nine miles of destruction
in its wake before being destroyed at close range by a howitzer. The
appearance of the T-34 in the summer of 1941 was a psychological shock to
German soldiers, who had been prepared to face an inferior ... 阅读全帖 |
|
w*******q
发帖数: 1764 | 39 DUSHANBE, Tajikistan (AP) — Tajikistan has agreed to give away a portion of
its territory to neighboring China
in a bid to put an end to a land dispute dating back more than a century.
Parliament voted Wednesday in favor of giving up around 1,000 square
kilometers of land in the Central Asian
nation's sparsely populated Pamir Mountains region. There was no immediate
information on how many
people live in the territory to be ceded.
Opposition leader Mukhiddin Kabiri said the land transfer is
unco... 阅读全帖 |
|
n****o
发帖数: 879 | 40 发信人: winetricks (winetricks), 信区: Military2
标 题: 俄罗斯称中国取代美国已成为最大的假想敌，要在远东遏制中国
发信站: BBS 未名空间站 (Wed Feb 9 10:15:39 2011, 美东)
Officially, Moscow denies that China is a potential threat, and endeavors to
play the political game of promoting friendship with Beijing. As a result,
any suspicion of China is excluded from Moscow's public discussions on
security and military strategies.
But the truth is that cracks have started to emerge in the seemingly solid
ties between the two powers since around 20... 阅读全帖 |
|
n****o
发帖数: 879 | 41 发信人: winetricks (winetricks), 信区: Military2
标 题: 俄罗斯称中国取代美国已成为最大的假想敌，要在远东遏制中国
发信站: BBS 未名空间站 (Wed Feb 9 10:15:39 2011, 美东)
Officially, Moscow denies that China is a potential threat, and endeavors to
play the political game of promoting friendship with Beijing. As a result,
any suspicion of China is excluded from Moscow's public discussions on
security and military strategies.
But the truth is that cracks have started to emerge in the seemingly solid
ties between the two powers since around 20... 阅读全帖 |
|
c**i
发帖数: 6973 | 42 (1) BNSF Railway
http://en.wikipedia.org/wiki/BNSF_Railway
Quote:
"The BNSF Railway (reporting mark BNSF), formerly known as the Burlington
Northern Santa Fe Railway, is an American freight railroad company
headquartered in Fort Worth, Texas; it is one of four remaining
transcontinental railroads and one of the largest freight railroad networks
in North America. Only the Union Pacific Railroad, its primary competitor
for Western U.S. freight, is larger in size.
"On November 3, 2009, Warren Buffe... 阅读全帖 |
|
d********a
发帖数: 3698 | 43 有一次极严重的核电厂燃料棒融化的污染！据估计其污染程度是三哩岛的457倍! 美国
官方一直到5
星期后才在报纸上悄悄的报道了这个事故.
“50 Years After America’s Worst Nuclear Meltdown
Human error helped worsen a nuclear meltdown just outside Los Angeles,
and now human inertia has stymied the radioactive cleanup for half a
century.
By Joan Trossman Bien and Michael Collins
http://www.miller-mccune.com/science-environment/50-years-after
meltdown-3510/
The Santa Susana Field Laboratory.
For Release Saturday A.M., August 29, 1959
CANOGA PARK, CA
“During an inspecti... 阅读全帖 |
|
u*******r
发帖数: 2855 | 44 Details of the death were sparse. He was killed in a military operation in
Afghanistan, according to sources, though it was unclear which countries
were involved. CNN reported that the al-Qaida leader was killed in a mansion
outside Islamabad, the Pakistani capital. |
|
C*******r
发帖数: 10345 | 45 【 以下文字转载自 USANews 讨论区 】
发信人: Commissar (柯宓同志), 信区: USANews
标 题: 纽约劳模女把三哥推下站台让地铁撞死
发信站: BBS 未名空间站 (Sat Dec 29 18:33:01 2012, 美东)
已经指控为谋杀和hate crime，哈哈哈。
泥鸽劳模真是未来模范统治种族。
NEW YORK—A woman accused of pushing a man to his death in front of a
speeding subway train in Queens has been charged with murder as a hate crime
, according to prosecutors, the Associated Press reported Saturday.
Police arrested Erica Menendez on Saturday after a passerby on a street
noticed she resembled the woman seen in a sur... 阅读全帖 |
|
c****h
发帖数: 4968 | 46 狗改不了吃屎的西媒，论调还是和奥运时一样：
“Unlike China, but like Europe, India expends a vast amount of its energy
simply coping with its incredible diversity. The French president Charles de
Gaulle once exclaimed: how can you possibly govern a country that has 246
varieties of cheese? Well, how about a country with 330 million gods?
China has diversity too, in vast if sparsely populated areas of mainly
Tibetan and mainly Muslim population, but it copes with it mainly by
repression.”
你玛，就是在给failed Indian democracy... 阅读全帖 |
|
R****a
发帖数: 6858 | 47 从而便于在中东北非发动针对美元的代理人战争.
The Russia-China Axis Grows
March 14, 2013 By Ariel Cohen 16 Comments
34 Print This Post
China’s new president Xi Jinping will make his first official foreign visit
later this month. He will visit Russia, in a trip Chinese sources say “
will improve relations and cement strategic partnership.”
Washington should pay attention to the strengthening ties between Moscow and
Beijing. These giant neighbors have the longest shared land border in the
world, and trade between the tw... 阅读全帖 |
|
l*****7
发帖数: 8463 | 48 1. 邪恶的日本独裁政府公然违反琉球王国人民的民主自由人权!
2. 强烈支持琉球王国的民主自由人权斗士!!!
3. 强烈支持琉球王国独立!!!
Check out link 1 and 2 below
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
Link 1:
http://www.nytimes.com/2013/07/06/world/asia/in-okinawa-talk-of
In Okinawa, Talk of Break From Japan Turns Serious
Ko Sasaki for The New York Times
Chosuke Yara, the head of the Ryukyu Independence Party, last month. “
Independence is an idea whose time has come,” he said.
By MARTIN FACKLER
Published: July 5, 2013
NAHA, Japan — In a win... 阅读全帖 |
|
N******K
发帖数: 10202 | 49 只有少数人搞生物的智商可以 比如什么sparse coding
很多都是弱智 混饭吃 |
|
z****e
发帖数: 54598 | 50 Rapping can be traced back to its African roots. Centuries before hip hop
music existed, the griots of West Africa were delivering stories
rhythmically, over drums and sparse instrumentation. Such connections have
been acknowledged by many modern artists, modern day "griots", spoken word
artists, mainstream news sources, and academics.[15][16][17][18] |
|
