Professional Documents
Culture Documents
Mat3004 17bit0296 PDF
Mat3004 17bit0296 PDF
Mat3004 17bit0296 PDF
DIGITAL ASSIGNMENT
REGISTRATION N0-17BIT0296
SUBMITTED TO-PROFESSOR PRAVEEN T
SLOT-C1+TCC1
APPLICATION 1-
- The color of the image can be expressed as three matrices R,G,B where Rij, Gij and
Bij are the amount of red, green and blue in the picture at position (i,j).
- A picture of size m*n can be expressed as the matrix of the same size. Suppose we
have 2*2 Matrix then the Rij, Gij and Bij can put together as
- The picture can be thought as an element of linear space with discretized co-
ordinates. The images can be scaled on various subspaces and make linear
combination. Various actions such as scaling, projection and the rotation can be
applied using the Basics of Linear Algebra.
- Consider the above two images where in both are manipulated from their original
state. By finding the dimensions of the given picture and putting it under
computation is bound to change the sub matrix of the original matrix of the image
expressing the Rij, Gij and Bij values.
- In terms of linear algebra while applying the filter it’s bound to change the each
pixel value of the matrix. As explained before, the input of this function can be just a
pixel like the adjustment of brightness, or a sub matrix of pixels like the blur, where
the order of the sub matrix will depend on the blur ratio.
𝑝11 ⋯ 𝑝1𝑁
[ ⋮ ⋱ ⋮ ]
𝑝𝑀1 ⋯ 𝑝𝑀𝑁
- When a linear transformation is applied on Matrix M containing the values for each
and every pixel of the image, a filter is basically applied onto the image. Here in this
assignment I have showcased two filters.
Here by applying this linear transformations image can be transformed according to the
needs of the user and this is how Linear Algebra is made use while applying the filters on
image.
APPLICATION -2
RECOMMENDER SYSTEMS AND
LINEAR ALGEBRA
- Recommender System predicts the likelihood that a user would prefer an item based
on previous user interaction with the data source that the system takes the
information from (besides the data from other users, or historical trends), the system
is capable of recommending an item to a user.
- Suppose there are n user and m movies. A few users have evaluated pictures, and
we have various triplets (i, j, y), where y is the rating given to film j by user i.
- For the matrix factorization way to deal with this issue, we will put the above data in
a rating framework Y. Y(i, j) is the rating user provided for motion picture j.
Nonetheless, Y is fragmented since each client has not appraised each picture.
Frequently, y is extremely sparse, with 1 of every 10,000 passages filled in.
- The goal is to assess the missing qualities in the matrix by finding a model f (i,j)=y
dependent on the accessible ratings, and after that to recommend to every user new
pictures which are relied upon to be most enjoyed by the user.
Y=U*𝑿𝑻
The first term measures the squared error between the actual and predicted ratings,
and the second term is a regularization term.
Once we use an optimization algorithm to find the values of U and X, the prediction
Y(i,j) is simply UiXj𝑇 .
- Singular Value Decomposition (SVD) is a strategy for finding the low-rank estimation
of a matrix, which is the thing that matrix factorization has been doing.
Decomposing a matrix with missing values is proportionate to setting the missing
qualities to 0, which will prompt unwanted outcomes. Henceforth to achieve better
outcomes matrix factorization method technique is ideal.
Question-
As the values of n and m are not equal in the question and contradicting the above
statement mentioned. Therefore the linear transformation would not be invertible.