#linear-algebra
2 messages Ā· Page 221 of 1
Assuming this channel is not in use
Is this the simplest way to calculate the cross product of 2 or 3d vectors?
Cross product is only in R^3
you can't cross a 2d vector with another 2d vector?
and I mean those are just the components of the cross vector
no
where's the 3rd dimension in R^2?
good point
so if a cross product is defined between two vectors P and Q, then both P and Q are 3-dimensional
Yes
why can't this apply for higher dimensions (i can't visualize it, anyway)?
Outer products exist for higher dimensional co-ordinate spaces
oh okay - i guess different types of products apply for higher dimensions
Yeah, you create inner product spaces by introducing an inner product like the canonical dot product
outer products just aren't covered in say first year linal
Does the screenshot I sent above essentially use the formulas for the determinants of matrices that are inside or part of a 3x3 matrix with the components of A, the components of B, and the components of the cross product of A x B
yes, the 1st row of the matrix is the unit vectors, then components of a and components of b
$a\cross b =det\left(\begin{bmatrix}i&j&k \ a_x&a_y&a_z \ b_x&b_y&b_z\end{bmatrix}\right)$
Mosh
right
to get the component of i, find the determinant made by the matrix that has 0 entries that share a row or column with i
cross out the row + column where i is in, and the matrix formed by the remaining entries is the determinant that becomes the component for i
yes, the cofactor
laplace / cofactor expansion is how you evaluate determinants
quick question: i've posted a linear algebra problem inside math help question, should I delete from there and eventually post it here so It is more "on topic" or those channels are for every math discipline?
sorry for disturbing though
doesnt matter
i think i'm done, but you're good š
ty mosh š
the help channels are for any question, higher level math tends to be put in the corresponding channel though
uh okay, so I'll leave it there. It's in #help-3 in case someone wants to solve a stupid problem I can't understand ahah
Hey how to calculate this, I assume it's with matrices ?
However h and k are not defined, so I'm not sure how to solve
multiply the top equation by -3
the values of h and k that work are the values h and k such that k = 3h
Ah okay I understand
Since the equations are equal after all
Only we need to divide the final k back to h based upon the left side of the equation and how they're related
yea. the image of the linear transformation represented by the given equations is a line, specifically the line y = 3x.
cool! not sure what ur second question is asking
I have no second question
I just had to find h and k which make the system consisten
oh. was confused on what u meant by this. nvm then
What do you call one "thing" in a matrix?
Like if you have the Matrix
[a b]
[c d]
a is a what of the matrix? Element?
There's various ways to generalize the cross product to higher dimensions, depending on what properties you want it to retain. The most similar but also pretty restrictive generalization exists in 7 dimensions: https://en.wikipedia.org/wiki/Seven-dimensional_cross_product
@edgy grail My function that I wrote to determine if lines intersect using linear algebra fails on partially overlapping parallel lines. The lines I tested were [(1, 1), (1, 3)] and [(1, 2), (1, 4)].
The reason it determined that the line segments did not intersect is because the matrix A was determined to be not invertible.
The matrix A being
[x1 - x0 x2 - x3]
[y1 - y0 y2 - y3]
Filling in the values, we get the matrix
[1 - 1 1 - 1]
[3 - 1 2 - 4]
Simplifying, we get the matrix
[0 0]
[2 -2]
Is it true that this matrix is not invertible?
yes, itās determinant is zero.
Interesting. The determinant is zero but the line segments still intersect.
well, thatās expected. theyāre the same line (the āthatā being zero determinant)
But they aren't the same line (segment).
hmm
I either forgot that or I didn't understand the significance.
Alright, thanks for the insight.
The line segment is on the line that that matrix describes, however.
Zero determinant means that the system might not have a solution, but it still could. A nonzero determinant means there will always be one unique solution.
You said your function is used to
determine if lines intersect
right? is that in 2 dimensions or 3
Because in two dimensions, unless they are parallel they will always intersect.
Do you maybe mean that the function finds where they intersect? Then the answer is not longer trivial
Zero determinant means that the system might not have a solution, but it still could. A nonzero determinant means there will always be one unique solution.
This is great information. I thought that zero determinant meant that the system will not have a solution. And that a nonzero determinant meant that there will be a solution.
right?
No. It determines if two line segments intersect.
is that in 2 dimensions or 3
That's in two dimensions.
Because in two dimensions, unless they are parallel they will always intersect.
I'm talking about line segments not lines. Non-parallel line segments may not intersect in two dimensions.
Do you maybe mean that the function finds where they intersect?
Currently, the function just finds the values oftands, the variables for the interpolation/extrapolation functions for the two line segmentsa(t)andb(s). I use these values to determine if the point of intersection lies on the line segments or on the infinite extension lines (extrapolated). The commented part of the following code sample explains how I usetands. (The comments start with a hash character)
# 0 <= t <= 1 describes all points on the first line segment
# 0 <= s <= 1 describes all points on the second line segment
# Therefore, if the point of intersection is on both line segments, then the line segments intersect
return 0 <= t <= 1 and 0 <= s <= 1 # True if intersect, False otherwise
Here is a sample of the code. It's pretty Englishy so if you don't know programming, you might still understand it.
Here's the part that is logically messed up.
# Matrix A must be invertible
A = Matrix([[x1 - x0, x2 - x3], [y1 - y0, y2 - y3]])
if not A.invertible:
print("A is not invertible so lines do not intersect")
return False # Lines do not intersect
The reason why I check if matrix A is invertible is because if it isn't invertible, then the determinant will be zero. Zero determinant is a problem because I need to divide by the determinant later on.
# The inverse of a matrix, A. is defined as follows
# [a b]^(-1) = 1 / (determinant(A)) * [d -b]
# [c d] [-c a]
# Therefore, the inverse of A for our system can be written as
# 1 / (det(A)) * [y2 - y3 -(x2 - x3)]
# [-(y1 - y0) x1 - x0]
If the determinant is zero and I divide by it, then that's undefined.
got a dumb question. the vector space C over R. how do you represent matrix multiplication?
like a matrix A would have real entries, but what dimensions would it have, if i want to do Av for some complex vector v?
I might be missing something.. but complex numbers arent matrices...?
Not with that attitude

@teal grotto If you're just considering C as a 2-dimensional vector space over R and identifying x + iy with (x, y) then I don't think you have a natural way to represent complex multiplication
But if you identify x + iy with a 2x2 matrix (x, -y; y, x) then you can check that complex multiplication corresponds to matrix multiplication
Additionally with this identification you get that |z|^2 corresponds to the determinant
I mean my prof had mentioned stuff like polar decomposition was like exp form of a complex number, but never understood what he meant lol
yea. maybe i didnāt explain that well. just want to have matrix multiplication with vectors in C
like if i want A to represent some linear transformation from C to C
Oh matrix vector multiplication?
i just want to do Av
ok yeah makes sense now lol
I see, line segments. That's interesting. I'd have to think about that before I could help any further.
@teal grotto not sure what you mean. like A is real and v is complex and you want to represent Av with all real matrices?
um. so R^2 as a vector space over R. itās two dimensional and linear transformations from R^2 to R^2 can be represented by 2 x 2 matrices with real entries.
asking for the analogue in the vector space C over R
Well you can have v'=(Re(v),Im(v))
C is isomorphic to R2
If that doesn't answer it then I think i still don't understand your question
oh wait. did i just not realize
C is a two dimensional R vector space. isomorphic to R^2
bruh
see. dumb question
Yeah
lol thanks
Np
@wintry steppe this is not linear algebra
Then what is it?? Where shall I post
Alright thanks
meguuuuu
and its det is 1 as well, would this work?
presenting this as an example of "det(A) = 1 implies A is a rotation matrix" does not constitute proof
just because this particular matrix of det 1 is a rotation doesnt mean they all are
and in fact,
$\bmqty{5 & 7 \ 7 & 10}$
Ann
hm okay so how do I approach this question then.....
well is the statement true or not?
well what about the matrix Ann wrote out?
um isee it has determinant as 1
and that its transpose is itself?
im confused what I am supposed to look for there...'
See what the matrix does to the basis vectors (0,1) and (1,0). Can the matrix possibly be a rotation?
ah no thats not rotational matrix because that matrix * its transpose is not identity matrix
Correct
Ah so that is one counterexample then
What I was trying to get at was unit vectors are not mapped to unit vectors
all i need to do is find the derivative of the equation
and then plug in 3 with all the values of x
quick question
if im correct this is independent right for the last row its 000 and there are two pivots remaining
nvm its dependent free variable
oops
Hey how to find this ? Manually I can only figure out one value for h
Nevermind
h != 2
Because they will be parallel at h=2
So, Riesz's representation theorem works for just inner product spaces, right?
Or is it solely for Hilbert spaces?
I think it's l.i. iff it reduces to the identity
"consistent system" just means there's one and only one solution
Oh the book says one or more solutions
However what are they asking ?
The question confuses me
Most notably, if you have a row that's like 0 0 0 a with a ā 0 it's not gonna be consistent
nope
also probably better in #advanced-analysis since it involves Hilbert space
Do you really think a+b+c=0 has no solutions because the RHS is 0
No I think g,h and k may not be zero ?
why couldn't they?
So, a nondegenrate inner product doesn't establish an isomorphism between a vector space and its dual?
it's aight
But what do they want from me?
I have the answer on my right but I don't know what it means
It shouldn't be problem for real or complex vector spaces, right? Since they are complete?
So, in these questions, you want to make a row of the form (0 0 0 something) appear
And it is possible, by doing row operations
If it wasn't, there would be no question, any values of g,h,k would do
Why do they want 0 0 0 if there are only 3 rows of the left size ?
what?
I learned that in trinomial like that in the question
And have three of them
They need to end up like
1 0 0
0 1 0
0 0 1
It is not a problem for the vector spaces R^n or C^n considered in linear algebra, but it is a problem for general vector spaces like the continuous functions on an interval example (with L^2 inner product) given in the stackexchange link. Note C(I) is complete using the sup norm but not with the L^2 inner product.
So why would we want 000 ?
You're mixing up some things
If you're able to get to the identity (the matrix you posted), then great
That means that your system is invertible i.e for any g,h,k, there's one and only one solution to the system
But that's not always the case
I don't know how to get to the identity in this one, because it contains variables
variables can be added and multiplied like anything else
but the thing is
you're not going to get the identity
you can try, but at some point you're gonna end up with a row of 0s whether you like it or not
In this scenario ?\
yup
Let's go over this again
if you could get to the identity, that would mean the system is invertible
any g,h,k gives you one and only one solution
in particular, you always have at least one solution
Who cares about the jargon, call it "always one solution", call it "invertible", call it "i can get to the identity", it's all the same
Do you really think that they would ask you this question (when does the system have solutions) if you could prove that the system always has solutions?
Actually, either it has none, either it has infinitely many
Okay that I understand
Usually, you'd expect three equations of three variables to completely determine a point
I mean, you should have enough info, right?
I guess, but it asks me to find an equation or something
we'll get there
If you can derive one equations from the other using linear combinations, well then that just means one of your three equations doesn't add any new information
Okay shall I do that now ?
Yeah, find which row can be expressed in terms of the others
I'm stuck haha
I can't find anything in here to merge
Is this some other starters have trouble with too ?
For example, if you're trying to see if the 2nd row can be expressed in front of the others, you first gotta at least make sure that it works for the first coefficient
The question would be then to find a and b such that a(-1)+b(2)=0
give me some
What do you mean with expressed in front of the others ?
I wish the book actually explained what they're doing instead of just showing the answers
Yeah...
If we call your rows R1, R2 and R3
Then we're trying to find, say, a and b such that R3=aR1+bR2
If this works for the entire row, it has to work for the first coefficient
Then you see if the numbers you found work for the rest
Okay and a and b will be used for the rest too ?
Which book are you using sentineL?
For the rest of the entries in the row
I just want to learn linear algebra, if it's this book or another, I don't care
This one is pretty good for computations I feel like
I came a bit late
So what was ur concern exactly?
Is there somethign specific you don't understand?
So just get it into reduced echelon form (remember that g,h,k are numbers just like the other ones)
Syst3ms do you mind if I ask a question really quick?
go ahead
So
it took me about a day of reflecting
But I finally understand where dot product comes from
this is what my book says
I understand that it works for perpendicular vectors
because of similarity
Btw if this is easy and I'm stupid, please say, so I can switch away from preparing for CS and pick a different career, I feel very dumb
nah, it's not trivial
Tim O'Brien
But this similarity is when these vectors x and y are perpendicular
How can a general definition emerge from this?
let me get my material on inner products
I found it h = -2g-k
Tim O'Brien
Like understand that the first equation comes because they are perpendicular vectors
And that has to be 0 otherwise you get 0=something nonzero
And this just proven that they're equal ?
So how do we make this jump with certainty to the general equation for all 2 vectors?
(sorry for the interjection)
if you take -2R1+R2-R3, you get a row of 0s and -2g+h-k at the end
So, if you don't want that to be impossible, your RHS has to be 0 too
Okay cool, I'm gonna read back and try to understand what I just did haha
Well, I feel like the definition of the dot product isn't so much "=0 for perp vectors" as much as it is "a measure of how much two vectors point in the same direction"
hmmm
and that's what is troubling for me
But it's ok
I just need to stop being a perfectionist around this and understand that some things I will understand later
Im just losing time sitting here thinking for hours
nah
honestly this doesn't feel like something that requires a lot of prior knowledge to discuss
I'll have you note a couple things
ok
First of all, there is little actually special about the (canonical) dot product
Last question, why are we trying to find that ?
$x_1x_2+2y_1y_2$ is also a valid inner product
Syst3ms
the problem is most easily seen when you have one row of 0s, you can immediately go "oh wait, the RHS has to be 0"
So, if you find something like that, you can just subtract rows to get your row of 0s
another thing of note is that given an inner product you can define the norm
as $|x|=\sqrt{(x|x)}$
hmm
Syst3ms
general notation for an inner product
the dot product is an inner product
the dot product has interesting properties, and inner products generalize that
What I showed is also a valid inner product, it just gives rise to a weird norm
Yeah yeah that's the pythagorean theorem
the length of the vector is just sqrt of its sides squared
So that's where dot product comes from ?
Honestly, I don't know enough about its history to tell you that
well I appreciate your help
But it is true that there are many possible inner products, but that only one of them gives us distance like we expect
hmmm
namely the dot product
say what now?
i dislike that notation of the dot product with itself, but yes
Tim O'Brien
$x\cdot x = x^2 = ||x||^2$
Tim O'Brien
but if you do, x^3 doesn't make sense
Ok
so i would refrain from writing x²
In the infinite dimensional case, it depends on the completeness of the space, yes?
I think what confused you is the order of the material. They didn't "go from" the one equation to the other. They said that we know two vectors are perpendicular iff x1y1 + x2y2 = 0. So this quantity x1y1 + x2y2 is of interest because it can tell us when vectors are perpendicular, so let's call that quantity the dot product. @forest quiver
(and on top of that it fits with the common notion of distance)
So the x1y1 +x2y2 = 0 thing was just motivating why we care enough about this quantity to give it a name
But how does that dot product work
when the two triangles don't have 90 degree angle
there
are they still similar?
is the question
I think they are right
no wait
thats dumb
A common geometric interpretation of the dot product in the general case is to orthogonally project one of the two vectors onto the other, multiply their norms, and add a negative sign if they point in opposite ways when projected
yeah I read about that on stack exchasnge
But whaaaaaaaaaaaaaaaaaaaaaaat
I mean it makes sense that its a projection
because 90 degree will yield 0
but idk
im not gonna worry about that
i cant get stuck on details like this
thank you tho
The reason why it's orthogonal projection is because the dot product of two perpendicular vectors is 0 :P
(you can prove this yourself, if you are convinced that for any vector x, you can split any other vector into one component that is aligned with x, and one that is orthogonal to it)
me neither
Micheal Penn
I just need to udnerstand where this damn projection comes from
Some cosine stuff?
What's a "signed angle"
it means it's positive if it points in the same direction, negative otherwise
@forest quiver Did you understand the book I am using in the first place ? These questions at the beginning of the book are so damn hard, after only reading 4 pages of linear algebra they ask me to calculate temperatures over a surface using linear systems
Ok right
those are the only 2 options you have, since they specify it has to be parallel
I understood the book well
it's pretty straightforward
You just learn the computation
and apply it
What things did you do in advance ?
I feel like I'm not prepared for this book
Ah okay yeah, I only did calc 1, that might explain it
No not rly
you dont need calc
to do it
just learn the computation and follow the steps its actualyl a great book. The one im doing now combines linear alg with calc 3 tho
They barely explain anything except how to solve an system, and then just throw a bunch of questions requiring deep understanding
Yeah this one
I loved this one
it took me like 30 mins to solve
You just follow the instructions
Did you find them hard to do ?
The second one you sent isn't hard
let me think about the first one
Consistent for all possible values
Yeah not too hard
it has at least one
They mean that it creates consistent system for all values of g and f
And ask what that says about c and d
The one with the temperature tho was very nice
It helped me understand it 10x better
but since it's for all possible f and g it might as well be asking that the system be invertible
Okay I should first do the temparatures one and then the others ?
well if c=1 and d=3 then its not consistent because g neq f
Yeah just create the equations
and line em up
it's more general than that
for this
(in general, when can you cancel out an equation by linear combinations)
yeah but we dont know the coefs.
Yes, otherwise you can take the completion and find linear functionals from there that can't be represented using the non complete space.
hmmm
Okay, gotcha! Thanks!
multiply the first equation, say, by a and see if you can find c and d that make the system not consistent
I now understand it indeed better
Linear algebra is useful when you're calculating stuff which depends on each other
In a circular fashion
Normal calculations would create infinite loop right ?
1 depends on 2 depends on 3 depends on 4 depends on 1
It's good when you have multiple linear equations
with multiple variables
because there exist simple algorithms
to isolate variables
and solve with ease
no
Fuck
I made an arithmetic error somewhere
These matrices are turning me crazy haha
This problem was so good that I have my sheet of work for this from like 2 weeks ago
lying on the floor
of my room
yes me too
Try it again
but this time have like a routine
Or I'm just using wolfram alpha, since it was just a arithmetic error
Because this part is correct right ?
Finally it's fun when I grasp it (kinda)
Not really
I haven't done it yet, found it too confusing
However I think that they need to be opposites of each other
In order to match at each f and g
-x+y
x-y
Forgive me if I'm wrong
Okay.
Linear algebra?
How do you usually solve ax +by = f, ca +dy = g using matrices?
What matrices will give consistent answers?
row reduce the augmented matrix [\begin{pmatrix}x & b & f \ c & d & g\end{pmatrix}] and back-substitute
Namington
(the x instead of a may seem weird, but it's because you wrote ca + dy = g instead of cx + dy = g)
if your a, b, c, d, f, g values are actually fixed (so x, y are your variables) then the answer is "you dont solve it using matrices"
you use ac + dy = g to solve for y
which is middle school algebra
and then plug that into ax + by = f and solve for x
you COULD use matrices but why overcomplicate it
you have a linear equation with 1 unknown
solve that
I'm trying to write some notes on Dual Spaces. Anything obvious that I missed, notation wise?
(There's been mention in this chat about covariant and contravariant elements, i'd like to work that in but i don't know how )
,rotate
v+W={v+w : w in W}
coycoy do u by any chance know if the question i linked is D and not A
V/W is the quotient space of V mod W and consists of cosets of the the form v+W. the cosets are the vectors in V/W
let me look
thnk u!
yes d looks correct
i saw that. i do not know fourier series. sry
coycoy linear algebra god
nah nah. thats TTerra
no problem
is there more than one correct answer to this? because a is correct too
fourth column is 4(1st) + 1(2nd) - 2(3rd)
ok, go ahead
i think the layout is correct
im just not sure where
the cos and -sin goes
like the layout
rotations are not really something iām too comfortable with yet. iām sure somebody else could help tho
thank u š¦
If someone can verify that would be amazing
This is what Iām thinking so
i get that if beta is a basis of V and T(beta) is also a basis of W then that would work for gamma=T(beta) but idk the requirements on beta
I wish I knew. I don't know that stuff yet š¦
no problem ;/
Integrate f(x)e^(-in2pix) to get the nth fourier coefficient
If you want to know how to do that, the hint is to split the integral from 0 to pi and from pi to 2pi
like
im confused
in2pix?
and why e^
like i know the integral of the first part from 0 to pi is
$in2*\pi*x$
saketh
So on the whole the integral is pi, correct. That will give the first coefficient
Also the way I was talking about fourier series used complex numbers, if you donāt do it that way you can ignore it
But you must have some formula for calculating the coefficients, right?
What about the other coefficients?
You divide all coefficients by 2pi
i thought for the bk terms u divide only by pi
What you divide by depends on how you define the integral
When you use e^(inpi) stuff you just divide by the length of the interval always
hmm ok
ig my question is
how can i consolidate the two functions into one
if even possible
i think that would make it easier
Yeah so here is the thing the integral of any function from to 2pi is the sum of the integral from 0 to pi and the integral from pi to 2pi
And inside of those two intervals f is just a constant, it would be -2 in the first case and 3 in the second
you're supposed to get a function of x
after doing the integral, you take those coefficients and get a sum of sines and cosines
No -2 and 3 are not the values of the integral, they are the values of the function f. Put those values in f and then do the integral, this should let you do all the integrals for the various coefficients
wait im so confused
wym put them into function f
f(x) is literally defined as f(x) = -2
and f(x) = 3
I mean instead of f write -2, that is all
right, but only for specific values of x
so integral of -2
so each of the integrals for the fourier coeffs has to be split into two
and integral of 3
ah ok
so
integral of -2 from 0 to pi
and integral of 3 from pi to 2pi
when i do that
i get
-2pi for the first
and 3pi for the second
integral of -2 * cos (...) from 0 to pi
and similarly for 3 * cos(...)
and also sine
what should go in the cos though?
what saketh told you above already
Yes those are the integrals for the first coefficient, iām saying you can do the same for the rest
ah ok
so
i know ak is = 0
so now im trying to find bk
there shouldnt be any cos terms in the answer i think
i think i got the answer can i verify w u @weak needle
or @lavish jewel
you can very yourself
plot the sines and cosines and see if it looks like the original thing
just post your working and iāll see if i can spot any mistakes, Iām a bit lazy too work out the integrals myself
cos(kpi) is not always -1, it will depend on wether k is odd or even
in which step
This is in the calculation of bk, sorry I forgot to say that
what should it be instead then?
Well, what is cos(kpi)? Just take a few examples and youāll see
is it k?
No
its either 1 or -1
Yep, and if you know k, how can you figure out wether cos(kpi) is 1 or -1
well
in the first one
i think its 0 to pi
so
it must be -1
and in the second one
its pi to 2pi
so it must be
1
Look, the stuff about the integral doesnāt matter here, all Iām asking is, given some integer k, what is cos(kpi). The answer depends on what k is, and what Iām expecting you to tell me is what the relationship is
when k is odd its 1 and when its even its -1
Exactly
So now do you see how the formulas for bk will change depending on wether k is even or odd
yeah
so then i think for the first one
its 2/pik
and then itll be
either positive or negative depending on
the order
and same w
the second
itll be 3/pik
and then depending on the order itll be positive or negative
does that sound fine?
No, if cos(kpi)=1 for example, what is cos(kpi)-cos(0)
And 1-1=0
Yes
so then for the odd ones its
-1 - (-1)
wait
it eliminates odd
and keeps even
bc i think its -1 and then the cos(0) is being subtracted
wait no
ur right
Yeah so the even terms are -4pi/k
Sorry -4/pik
just for š(š„)=ā2
and so for the second part its
and so final answer i think is
Why is it 2/3pi and 2/5pi and so on? That seems wrong.
First of all, all the odd k coefficients should be 0 right?
oops
No iām sorry i think your rightp
Yes you were right, I got it mixed up
Well now that you got the steps, I think it is a little pointless to verify the addition or whatever, just do what edd said and plot it on a computer or something
(Sorry Iām a bit lazy)
u can plot it?
There should be graphing calculators available online
thank u saketh for ur help and patience i really appreciate it
No problem
for a0 the first coefficient
should u integrate from pi to 2pi
or 0 to pi
Pi to 2pi
Well the integral from pi to 2pi of 3 is the same as the integral from 0 to pi of 3
Np
does anyone know how to solve this?
try diagonalizing A first
just in case, A should be diagonalizable over R (checked because its characteristic polynomial has two real roots, 2 and -2, which are the e-vals of A).
once you diagonalize it, you will have a diagonal matrix X and an invertible matrix P so that A = P X P^-1. then A^6 is P X^6 P^-1 and youāre done, since D = X^6 is diagonal
Hey, when having this
It is wanted that the pivot is the most right right ? (2, 3)
When it's the last pivot in the augmented system
well, it's just a sequence of numbers (or something else)
you have a first value, then a second value, then ...
And I don't understand why they say it's a vector space
It just doesn't make sense
oh, that
Well, first of all do you know that the set of functions from ā to F is a vector space?
I don't get what you're saying
Do you even know what a vector space is?
a sequence can be considered a vector in a vector space
I read it a lot of times. I don't get what it means
set of vectors multiplied by scalar?
vectors with inputs from a field F?
its a set of vectors
Here's how I understand it
simply put
A vector space is a set, along with some given field F. The elements of that set are called "vectors".
They're very much analogous to the vectors you see in geometry/physics
in fact, they're a generalization of that
Yes I know. But here you can have n tuple array vectors
kinda like vector in n dimensions?
yeah?
so long as you define + and scalar multiplication well, why would it be an issue?
So do I know what vector space is?
Well, there are some additional properties that vector spaces need to meet
basically, adding works well
you can scale things, and that works how you'd expect
Yeah
scalar multiplication isnt necessarily a vector space axiom, it's just part of defining the * operation from VxK to V
also a vector space has to be a group for +
group?
$\cdot :V\cross\mathbb{K}\to V$ is scalar multiplication
I'm confused as to how you got to vector spaces without knowing what a sequence or a group is
Mosh
group theory comes after linal usually
well that's dumb and stupid
I know what sequence is. Didn't understand what it meant in LA
same thing in like calc / other settings
can confirm
${a_n}=a_1,a_2,a_3,...$
Mosh
anyway, group for + just means that there is something like 0, and every element has an additive inverse
(which you'd expect vectors to)
Can't you just say existence of inverses?
I mean existence of inverses and every element has an additive inverse technically mean different things
since you can have operations which define only certain elements having inverses
Some of the vector space axioms are really stupid things, like $1\cdot x = x$
makes sense
Syst3ms
(if you multiply by 1 it doesn't change anything, duh)
by stupid i mean stupidly obvious
depends on the identity and element
as long as you call 1 the multiplicative identity like any reasonable human being, this must work
Commutativity of +
Associativity of + and *
Existence of additive and multiplicative identities
All elements have an additive inverse
- and * distribute
those are the 8 axioms in few word basically
Now bear in mind, as long as all these axioms work, things can be n-tuples, but they can very well not be
In fact, that's the point : going beyond tuples, and looking at the concept in its full generality
So now, back to the main point
Why is the set of functions from ā (or any set) to F a vector space?
Well, you can define function addition by (f+g)(n)=f(n)+g(n)
You can define scalar multiplication by (λ·f)(n)=Ī»f(n) (recall, Ī»āF)
And it isn't hard to see that all properties mentioned before hold
By definition, this set is a vector space (over F, at least)
But isn't a vector space about representation of n tuples? I don't get when a function is a vector space
a set is a vector space right?
How is a function one?
?
Vector spaces are about generalizing the operations and concepts that notably apply to n-tuples
In this case we're replacing the n-tuples with functions
just a side note on identities maybe im stating the obvious but in case you miss the point
identities are operations that keep the element of the set as is
so even something as trivial as 1* x=x 1 being the identity for multiplication
the identity of lets say a matrix A with the multiplication identity would be A*(the identity)=A that you would know as In the identity matrix
and the identity of a set M with the union would be M U (identity)=M thus being phi
ik your topic is vector spaces but had to point it out
formatting
ye discord has this thing for *
How can you give functions as inputs. I don't get what you're saying here
And why doesn't this book explain what's going on
Inputs of what?
You said you're replacing the array of n tuples with functions
why isn't there anything explained in the book?
I wouldn't know
i don't quite get what you think i am implying by that
let 'f','g','h' be a function. Instead of a vector x =(a_1,a_2,.....) didn't you say a vector x can also be =(f,g,h)?
No, that's not my point
My point is that you know how we can add and scale vectors?
x=(N,F)
No..?
yes we can add vectors
Well we're defining an analogous way to add and scale functions
Which works in a very similar way to vectors
But the two kinds of objects have nothing to do with one another
Just so we're clear
The book sucks
anyway, addition and multiplication of sequences is the same thing
just with different notation
Though, one last tidbit that was in the original definition
you can always use another book for reference, helps
It specifically considers sequences that are nonzero for a finite number of terms only
Bruh then what's the point of the original book
I sincerely thought I can be done with LA in 6 months without hassle if I stick to one book
yeah but some books aren't good teachers
heck, we don't have textbooks here
i mean, in highschool we kinda do but we mostly use them for exercises
i digress
I'm using Friedberg,Insel,Spencer BTW. his definitions are marked as "Example:n"
well you can
1-complain about the book and sit in a corner
2-think reasonably and get another book that will solve the lack of competency from the book
you digress with?
digress, not disagree
my bad
anyhow
Yeah
this
Yes. On the search now
we still haven't proved that this was a vector space
we proved that the set of functions from ā to F is
But not with that restriction
When you have a vector space that is a subset of another vector space (with the same operations), you call that a vector subspace (how creative)
Checking if something is a vector space is tiresome
You have like 8 axioms and it sucks
It's much faster to see if something is a vector subspace of a vector space you already know
I chose to start by explaining how the set of functions from ā to F was a vector space because it's nice to keep in mind, since it's really general
Checking if a set is a vector subspace is easy on the other hand
You just have to check :
- the set is included in the original vector space (duh)
- 0 is in the set
- the set is stable by scalar multiplication
- the set is stable by addition
i think you won't have trouble checking that the first two hold
the set of sequences that are only nonzero for a finite number of terms also includes the sequence which is always 0
it's stable by multiplication because you're not going to change any of the zero terms by multiplying them
And it's stable by addition, because you're still going to end up with only a finite (but perhaps larger) amount of nonzero terms
So there you have it, the set of sequences with a finite amount of nonzero terms is a vector space

I..... umm
take your time
I don't understand when you say "a set of functions" is a vector space
How are functions depicted as vectors?
a vector is no longer necessarily just an n-tuple
it's an element of a vector space
that sounds circular i know (and it kind of is)
but really, a vector in the most general sense is only defined by how it transforms with regards to addition and scalar multiplication
does it behave well and expectedly with regards to those two?
If so, it can be considered a vector

maybe I'll try tomorrow
after some headbanging
I heard this subject is supposed to be easy
it isn't all that bad, but there are some initial subtleties
main take away from this is that a vector is no longer just an n-tuple, it's way more general now
well it is once you get the hang of it
its not as hard to grasp you'll be fine

I have a question. If I have an augmented matrix and almost all the numbers are filled in, but there is an unknown coefficient in the matrix how do I solve for it.
depends on what you're asked for
are you asked to find h such that the system has no solutions?
or are you asked to find h such that the system has (420, 69) as a solution?
or what
@dusky epoch has infinite solutions
okay
so when a system has infinite solutions, row-reducing it will eventually lead to a row of all zeros
act accordingly
An augmented matrix is inconsistent when a row of zeros leads to a number right?
if two rows are exactly the same
then the system can't be inconsistent
theres no such thing as an augmented system
the system can still be inconsistent if two rows are the same
consider the augmented matrix $\begin{pmatrix}1&1&0\1&1&0\1&1&1\end{pmatrix}$
Namington
Sorry i should've specified more
this row reduces to $\begin{pmatrix}1&1&0\0&0&1\0&0&0\end{pmatrix}$ which translates into the system $x + y = 0, 0 = 1, 0 = 0$
Namington
and obviously 0 = 1 is a nonstarter
jswatj
not necessarily, consider $\begin{pmatrix}0&0&1\0&0&1\end{pmatrix}$
Namington
that said, "most of the time" yes
in that if a or b are not equal to 0
OR if the rightmost column is equal to 0
then yes, it will be consistent
that is to say, the ONLY inconsistent matrices of that form look like $\begin{pmatrix}0&0&k\0&0&k\end{pmatrix}$ for some $k \neq 0$
Namington
hmm ok
("inconsistent matrix" here meaning augmented matrix representing an inconsistent system)
right
unless a = b = c = 0 or e = f =g = 0
then the system isnt inconsistent
so?
š
"if these are inconsistent, then theyll be parallel" doesnt mean "if these are parallel, then theyll be inconsistent"
all dogs are animals, but not all animals are dogs
the contrapositive would be "if these are not parallel, then they are consistent" - which is actually true!
(assuming they form planes, at least)
if your planes arent parallel they'll cross at some point
(more appropriately, some line)
the line on which they cross is your solutions
right
so its true
to form a plane you need all your coefficients to be nonzero
yeah
i added that but it was redundant in hindsight
so forget that bit
basically, if we have two planes, then there are these possibilities:
- The planes are not parallel, meaning they cross at some point [CONSISTENT]
- The planes are parallel, meaning either:
--- They are "the same", so they intersect at every point [CONSISTENT]
--- They are "stacked over each other" but never meet, so they never intersect [INCONSISTENT]
not necessarily that the coefficients are the SAME
right!
one equation is a scalar multiple of the other (except the constant term/rightmost column of the augmented matrix)
sure
Does Riesz's representation theorem work for inner products that are not positive definite?
that's a contradiction
can you get an orthonormal basis for such a space
"inner product" literally implies positive definite
but it wouldn't work if it wasn't positive because you need cauchy-schwarz
LOL, pseudo inner product.
and at one point the proof uses the fact that <a,z>=0 for all z implies a=0
It could, but there would be the signature involved.
which uses the definiteness of the inner product
and that's not even mentioning the more advanced theorems which may implicitly come from either of those two properties
what are the sort of theorems you are talking about?
im thinking if basic ones like triangle inequality, pythagoras, and gram schmidt hold
(and i dont rlly think so)
idk i just skimmed over the proof
Does results like Riesz's representation theorem exist for spaces with a pseudo inner product?
Otherwise, how can one use the "metric" to "lower" or "raise" the indices in an expression as they do in General Relativity where the Lorentzian inner product on the tangent space is a pseudo inner product?
all you need to raise and lower indices is non-degeneracy of the "metric tensor," no? and that's built into the definition
if you have a free variable after reducing a matrix into row echelon form, does that mean that the vectors that make up the matrix are linearly dependent?
yes
yeah, it's like saying positive definite matrices are the only invertible matrices, that's just not the case
Isn't this non-degeneracy of the inner product that is the isomorphism from Riesz's representation theorem?
Can I establish an isomorphism between a vector space and its dual with a pseudo inner product?
How do you define a topology using pseudo inner products to talk about continuity? Genuine question because I know nothing about pseudo inner products.
If there's some way to make a norm out of a pseudo inner product, I think you can; I don't think you can make a norm out of a pseudo inner product though.
So then how do you construct the dual space?
You don't need an inner product to construct the dual space.
You need to know what the continuous linear functionals on your space are
For which you need a topology to talk about continuity?
Isn't the dual space just bounded linear functionals?
How do you define bounded then?
And linearity implies continuity, IIRC.
