A general note: Please write out proofs as carefully as you can.
At any point where you use any result, either from class or from the text,
state the result carefully. I have given below a sample problem
that indicates the approximate level of detail you should give in a proof.
Example: Let V be a vector space over a field F. Let U, W be subspaces of V, neither of which contain the other. Then the union of U and U' is not a subspace.
Proof: Suppose, by way of contradiction, that the union of U and U'
is a subspace.
By assumption, there exists a vector u in U which is not in U',
and a vector u' in U' which is not in U. Let v = u + u'. Then v does
not belong to U, because, if it did, u' = v-u would belong to
U, contradicting our assumption. Symmetric reasoning
shows that v does not belong to U'.
We have exhibited two elements (u and u') of the union of U and U',
whose sum (namely, v) does not belong to the union. Therefore, it is not a subspace.
Problem set 1, due Friday October 3 at 5pm:
- Axler Chapter 1, problems 6, 7, 13, 15 (We will cover dimensions on Monday).
- Axler, Chapter 2, problem 1, 6.
- Let V be a vector space over the field R of real numbers.
Prove that V is not equal to the union of a finite number of proper subspaces.
(A subspace is called proper if it is not equal to V.)
- Let A be a list of vectors in a vector space V. Show that span(A) is the intersection of all subspaces containing A.
- (Extra credit) Can a vector space be the union of three proper subspaces?
Solutions.
Problem set 2, due Friday October 10 at 5pm:
- Axler Chapter 2, problems 8, 11,13, 14, 15.
- Axler Chapter 3, problems 1, 3.
- (Katznelson 1.2.7). Suppose that U, W are subspaces of a vector space V
so that U intersects W only in the trivial vector. Suppose
that u1, u2, ... un is a linearly independent list in U,
and w1, ..., wm is a linearly independent list in W. Show that
u1, u2, ..., un, w1, ..., wm is linearly independent in V.
- (Katznelson 1.2.9; previously listed incorrectly as Katznelson1.2.8) Suppose that V is a finite dimensional vector space.
Show that every subspace W of V satisfies dim(W) <= dim(V),
and that equality dim(W) = dim(V) holds only when W=V.
- Let V be a finite-dimensional vector space over the complex numbers. Let V'
be V BUT considered as a vector space over the real numbers.
Show that the dimension of V' is twice the dimension of V.
- (Extra credit). Suppose that we are given a system of linear equations
all of whose coefficients are integers (example: 2x+3y+4z=2, x-z=8)
and you know that it has a solution in real numbers.
Prove that it has a solution in rational numbers.
Solutions; another set of solutions by the CA
is here.
Problem set 3, due Friday, October 17. Because
of midterms, you can hand in any FOUR of the Axler problems
and ONE of the written problems below, although I recommend
that you try them all if you have the time.
- Axler Chapter 3, problem 4,6, 12, 14, 15, 16, 23, 25.
- We denote by R the real numbers
and Q the rational numbers. Let x in R. Prove that
the following two statements are equivalent:
- x is algebraic, i.e. ,
satisfies a polynomial equation with rational coefficients.
- Consider R, the real numbers, as a Q-vector space.
The subspace spanned by the infinite list 1, x, x^2, x^3, ...
is finite-dimensional.
-
Let V be an n-dimensional vector space over a field,
and let T: V --> V be a linear map. Let T^k
denote the kth power of T (i.e., T^2 =T.T = T composed with T,
T^3 = T. T^2 and so on). Suppose T^k = 0 for some positive integer k.
Show that T^n = 0.
Give an example to show that T^(n-1) need not be zero.
Problem set 4, due Friday, October 24. With apologies for the late posting ! - AV .
(These questions
are somewhat more abstract than what we have encountered so far. In all cases, if you don't know
where to start, first try to figure out an example; partial
credit will be given for any correctly worked out special case.)
- Describe (with proof) how the adjoint of a composition
of two linear maps relates to the adjoints of each linear map separately.
- Suppose X, Y are two vector spaces.
Let Z be the direct sum of X and Y. Prove that Z*
is isomorphic to the direct sum of X* and Y*. (Note:
do not assume the vector spaces are finite dimensional.)
-
In the following questions, V, W are finite dimensional vector spaces
over the field F, T: V-->W is a linear map
and T^: W*-->V* be the adjoint.
Prove the following.
- There exists a basis e_1, ..., e_n for V
and a basis f_1, ... f_m for W so that T(e_i) is either equal to f_i,
or equal to zero.
Describe the adjoint T^ with respect to the dual bases e_i*, f_j*.
- If T is surjective, then T^ is injective.
- If T is injective, then T^ is surjective.
- The dimension of the image of T and the dimension
of the image of T^ are the same.
- Suppose that V is a finite-dimensional
vector space with basis e1, ... en. Let e1*, ..., en* be the dual basis
for V*.
Suppose that T: V-->V* has a matrix a_{ij} (with respect to these
bases) that is symmetric, i.e., a_{ij} = a_{ji}. Prove
that this is "independent of basis", i.e., given any other basis
f1, ..., fn, with dual basis f1*, ..., fn*, the matrix of T with respect to (f1, ..., fn) and (f1*, ..., fn*)
is also symmetric.
(Hint: the elegant way to solve ths problem is to find a way
to phrase "symmetric" without reference to a basis.)
Solutions.
Problem set 5, due Friday October 31, 5pm.
The determinant of an nxn matrix M with entries in a field F is the
determinant of the linear transformation F^n ---> F^n given by matrix multiplication with M.
- Suppose that M is a square upper triangular matrix. (This means
that all entries of M strictly below the diagonal are zero.)
Prove, from the definition of determinants
as presented in class, that the determinant of M
is the product of the diagonal entries.
- Let D be a nonzero alternating 3-form on R^3.
Describe in geometric terms
when, for v1, v2, v3 in R^3, the sign of D(v1, v2, v3) is positive.
- Suppose that D is a nonzero alternating n-form
on an n-dimensional vector space V. Suppose that e1, ..., en is
a basis for V.
- Prove that, if we replace e1 by e1 + alpha e_j, for any alpha in F
and any j > 1, the value of D(e1, ..., en) remains unchanged.
- Prove that, if M is an nxn square matrix, then
adding any multiple of a row of M to some other row
leaves det(M) unchanged. (This gives an efficient way of computing the determinant;
if you have not seen determinants before you should compute some determinants of 3x3 matrices
using this property and Problem 1.)
- (This is Katznelson Corollary 4.4.4; try to prove it yourself
before looking it up.) Suppose that V is the interior direct sum of subspaces W1, W2.
Suppose T: V --> V "preserves" W1 and W2. This means that whenever w in W1, then also T(w) in W1,
and similarly for W2. Let T1 = T restricted to W1,
and let T2 = T restricted to W2. Prove that det(T) = det(T1) det(T2).
- Do either Katznelson IV. 5.5. (Hint: the prior problem may be helpful)
or Katznelson IV. 5. 12. (The problem about Vandermonde matrices.)
-
Suppose that M is a 2x2 matrix with entries a,b,c,d.
Suppose that, for every scalar x in F, the determinant of xM + Id
(where Id is the 2x2 identity matrix) equals 1. Prove that M^2 = 0.
- Extra credit: generalize the previous question to nxn matrices (partial credit for 3x3).
- Extra credit: Let A: V --> V be a linear transformation
on a finite dimensional space V. Let A^: V* ---> V* be the adjoint.
Prove det(A) = det(A^).
Solutions.
Problem set 6, due Friday November 7 5pm.
- Axler Chapter 5, problems 4, 5, 6, 7, 11, 12, 21, 23. (For problem 5,
you can look up the definition of "invariant" in the first section
of Axler Chapter 5).
- Suppose that the characteristic polynomial of a 3x3 matrix M
is det(x. Id - M) = (x-1)(x-2)(x-3). What is the determinant of M? What is the characteristic polynomial
of M^100?
Solutions
Problem set 7, due Friday November 14, 5pm. (Apologies for late posting!)
Axler Chapter 6, problems 2, 3, 5, 6, 9, 13, 21, either 22 or 23.
Extra credit: A variant of the problem from class: In R^n, how many
vectors can you find which are mutually obtuse? In other words,
what is the largest size of a list v1, ..., vk so that every
inner product of distinct vectors is negative?
Comment: This type of question is related to coding theory; can you see why?
solutions.
Problem set 8, due Friday November 21, 5pm.
Axler, Chapter 6, problems 24, 27, 29, 30.
Axler, Chapter 7, problems 1,2, 3(b), 11, 14.
Extra credit: Suppose that V and W are finite-dimensional inner product spaces.
Let L be the space of linear maps from V to W.
Describe a natural inner product on L.
(Natural means that it does not depend on any choices. Thus,
if you choose a basis in order to define your inner product, you should
later prove that it was independent of the choice of basis.)
Comment: This is a useful thing to know, because
we would like to be able to make sense of ``one linear transformation approximating another.''
solutions.
Problem set 9. NOT FOR ASSESSMENT (however, if you wish to hand
in any of it, it can be graded to see if it is correct!)
- Axler, Chapter 7, problems 30, 31, 32, 33, 34.
- In Axler problem 7-34, What happens if you ask not about
the largest and smallest singular values, but rather the second largest
or second smallest? What happens if you asked
about the singular values of the product of two matrices, rather than the sum?
- Write down a 3x3 matrix and compute, in horrid detail, its singular values and singular vectors. (This is well worth doing once, just to make sure
that you can turn the theory into practice.)
- For an example of how SVDs are useful in practice, look up "latent semantic indexing," and try to understand why the SVD helps (i) disambiguate
words which have different meanings in different contexts and (ii) relates
different words with the same meaning.
- Let V be a complex vector space and S: V-->V an invertible operator.
Prove that S has a square root: there exists T so that T^2 = S.
(Hint: use the Jordan normal form; this is actually proved in Axler 8.32
if you get stuck.)
Solutions.
Akshay Venkatesh
Department of Mathematics Rm. 383-E
Stanford University
Stanford, CA
email: akshay at stanford math