Linear Algebra - Lecture 16 - Fall, 2026
Overview
A square matrix is invertible when it has a two-sided inverse, which is unique; the 2 by 2 formula uses the determinant, while Gauss-Jordan reduction of the augmented matrix [A|I] computes inverses of larger matrices or reveals that no inverse exists. The discussion then derives a change-of-basis matrix: if each vector in basis F is expressed in basis G, its coefficients form a matrix P that converts coordinates from F to G.
Key takeaways
- A two-sided inverse is unique: the equations AB = I and CA = I imply B = C by associativity.
- For a 2 by 2 matrix, det(A) = ad - bc; a nonzero determinant allows the inverse formula (1/det(A))[[d,-b],[-c,a]].
- To compute an n by n inverse, reduce [A|I] using elementary row operations; a final form [I|B] identifies B as A inverse.
- The example matrix [[2,-3,0],[0,-1,4],[6,1,0]] has inverse [[1/20,0,3/20],[-3/10,0,1/10],[-3/40,1/4,1/40]].
- For bases F and G, the change matrix P has columns equal to the G-coordinate vectors of the F basis vectors, so [v]G = P[v]F.
Chapters
- An n by n matrix A is invertible if a matrix B satisfies both AB = I and BA = I.
- The identity matrix I has ones on its main diagonal and zeros elsewhere, and multiplying by I leaves a matrix unchanged.
- The inverse is unique: if AB = I and CA = I, then C = C(AB) = (CA)B = B.
- For A = [[a,b],[c,d]], multiplying by [[d,-b],[-c,a]] produces (ad - bc)I.
- When det(A) = ad - bc is nonzero, A inverse is (1/(ad - bc))[[d,-b],[-c,a]].
- For larger matrices, the lecture introduces an algorithm to determine invertibility and compute the inverse.
- For A = [[5,7],[2,3]], augment with I to form [A|I], a 2 by 4 matrix.
- Elementary row operations reduce the left block to I; the right block becomes [[3,-7],[-2,5]].
- The result matches the determinant formula, since det(A) = 5·3 - 7·2 = 1.
- If the left block cannot be reduced to I, the matrix has no inverse.
- The example starts with A = [[2,-3,0],[0,-1,4],[6,1,0]] and appends the 3 by 3 identity matrix.
- Row operations create pivots in all three columns of the left block, then eliminate entries above and below each pivot.
- The resulting inverse is [[1/20,0,3/20],[-3/10,0,1/10],[-3/40,1/4,1/40]].
- For an n by n matrix, the augmented matrix has n rows and 2n columns; when reduction yields [I|B], B is A inverse.
- In an n-dimensional vector space, a basis F = (f₁,…,fₙ) gives every vector v a unique coordinate vector [v]F.
- A second basis G = (g₁,…,gₙ) gives the same vector a generally different coordinate vector [v]G.
- The change-of-basis matrix P is defined so that [v]G = P[v]F.
- Express each fⱼ as a linear combination of basis G: fⱼ = Σᵢ pᵢⱼgᵢ.
- Place the coefficients pᵢⱼ in row i, column j of P; column j records the G-coordinates of fⱼ.
- If [v]F has entries xⱼ, then the coefficient of gᵢ is Σⱼ pᵢⱼxⱼ, giving [v]G = P[v]F.
- The vector v stays fixed during a basis change; only its coordinate representation changes.
Transcript excerpt
0:09 Now, this is lecture 16 in linear algebra. And we're going to discuss how to compute the inverse of a matrix. So, let's recall um that A be an n by n matrix. So, the number of rows equals the number of columns. So, this is what is called a square matrix. It's not a rectangle, it's actually a perfect square. Uh Uh A is invertible if there is a n by n matrix B
0:59 such that A * B is the identity and B * A is the identity. Where the identity matrix is the matrix um I can write the ijth coordinate as delta ij where delta ij is equal to 1 if i equals j and 0 if i is different from j. So, for example, I3 3 by 3 matrix you have ones down the main diagonal where i equals j and zeros everywhere else. I2 would be 1001. I4 would be 1000
1:49 0100 0010 and 0001. This is the identity matrix and the identity matrix of course has the property that any matrix by the identity just gives you back the same matrix. So, A is invertible if there's a matrix B where AB is the identity and BA is the identity, and B is called the inverse of A and denoted by A to the power minus one. And we can also observe something nice with this uh uh simple theorem, perhaps.
2:43 If A is invertible then its inverse is unique. That means there's only one inverse. So, suppose AB is the identity and CA is the identity. So, suppose B is an inverse of A and C is an inverse of A. Must show B equals C.
3:40 So, how do we show that? Well, C is C times the identity. Cuz any matrix multiplied by the
This is the start of the transcript. Save this video free for the whole transcript, searchable and time-coded.
In this course
Related notes
Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, Nathanson math lectures.