Quiz 知识点: - 原题问 "对于这类应用,我们关心大维度
和大数据集 ,其中 通常_",答案是 Much smaller
than。ANN 场景中 远小于
,但精确 NN 在高维下有指数级下界。
- 原题问 "ANN 放宽 NN 的方式是_",答案是 Allowing
to return a point that might not be the closest to the
query(允许返回可能不是最近的点)。这是 ANN
的核心思想:用近似换取查询效率。
[EN] We know how to solve the (exact) Nearest
Neighbour question over a -dimensional space with query time ____
and space ____.
[CN]
维空间中的精确最近邻查询,查询时间_,空间_。
选项
答案
❌
✅
❌
❌
知识点:链表方案同时满足这两个界——线性扫描所有 个 维元素。
Question 2
[EN] Typically, for this type of applications we
care about large dimension and
large dataset , where is _____
[CN] 对于这类应用,我们关心大维度 和大数据集 ,其中 通常____。
选项
答案
Much larger than
❌
Equal to
❌
Much smaller than
✅
Comparable to
❌
知识点:ANN 场景中 ,但维度诅咒仍然让精确 NN 不可行。
Question 3
[EN] Suppose the number of points is huge: exponential in the dimension
. The known algorithms for the
(exact) Nearest Neighbour problem have query time or space complexity
that scales ____ with the dimension .
[CN] 若点数
是维度
的指数级,精确最近邻算法的查询时间或空间复杂度随
选项
答案
Exponentially
✅
Logarithmically
❌
Linearly
❌
知识点:精确最近邻在高维下遭遇"维度诅咒"(curse of
dimensionality),复杂度指数级增长。
Question 4
[EN] The Approximate Nearest Neighbour problem
relaxes the Nearest Neighbour problem by...
[CN] 近似最近邻(ANN)放宽 NN 的方式是……
选项
答案
Allowing exponential space
❌
Allowing a probability of failure
❌
Allowing to return a point that
might not be the closest to the query
✅
知识点:ANN
的核心妥协:允许返回不是最近邻的点,换取查询效率。
Question 5
[EN] The Johnson-Linderstrauss Lemma allows us to
preserve exactly the Euclidean distances between points, but on a much smaller space of
dimension only .
[EN] The Johnson-Linderstrauss Lemma allows us to
preserve approximately the Euclidean distances between points, but on a much smaller space of
dimension only .
[CN] JL 引理可近似保持欧氏距离,只需 维。
选项
答案
True
❌
False
✅
知识点:目标维度是 (与点数相关),不是 。这个是最常见的混淆点。
Question 7
[EN] The Johnson-Linderstrauss Lemma allows us to
preserve approximately the Euclidean distances between points, but on a much smaller space of
dimension only .
[CN] JL 引理可近似保持欧氏距离,只需 维。
选项
答案
True
✅
False
❌
Question 8
[EN] The Johnson-Linderstrauss Lemma allows us to
preserve approximately the Hamming distances between points, but on a much smaller space of
dimension only .
[EN] Locality-Sensitive Hashing uses a type of hash
functions which makes collisions more likely when hashing elements close
to each other.
[CN]
局部敏感哈希(LSH)使用一类使相近元素更可能碰撞的哈希函数。
选项
答案
False
❌
True
✅
知识点:LSH 的核心理念——碰撞概率随距离单调递减。
Question 10
[EN] Using LSH, one can solve the ANN question in
Hamming and Euclidean space with expected query time _____ in the number
of points , and space _____ in
.