Implement polynomial curve fitting in python and with TensorFlow (optionally).
- sample the function curve of y=sin(x) with Gaussian noise
- fit degree 3 and 9 curves in 10 samples
- fit degree 9 curves in 15 and 100 samples
- fit degree 9 curve in 10 samples but with regularization term
- Programming lanuage: python
- Optional package: tensorflow web
- Plot the results in matplotlibIntroduction in Chinese Tutorial
Fit degree 3 and 9 curves in 10 samples:


Fit degree 9 curves in 15 and 100 samples:


Fit degree 9 curve in 10 samples but with regularization term:

Degree 9 in 10 and 100 samples using Tensorflow with regularization term (20000 iterations):


Represent digits '3' in 2D
- convert data from the UCI Optical Recognition of Handwritten Digits Data Set
- perform PCA over all digit '3' with 2 components
- plot the PCA results
- Programming lanuage: python
- Plot the results in matplotlib
- Use TensorFlow if possible
First digit '3' in the dataset:

PCA over all digit '3' with 2 components:

Raw digit '3' on red dots above:

Visualization of mean and first 2 components of digit '3':

3. MOG
Implement MOG in 2D case
- Generate sampling points from a 2D Gaussian distribution
- Generate sampling points from a 2D Mixture of Gaussian (MOG) distribution
- Estimate parameters of generated data via step2 by using the E-M method
- Plot iteration results
- Programming lanuage: python
- Plot the results in matplotlib
Using EM method to learn parameters of 2D-MOG distribution:

Comparison of estimated (dashed) and actual (solid) values:

4D-MOG using EM method:


Implement the Levenberg-Marquardt method
- Programming lanuage: python
- Design your own testing functions
5. SVM
Implement (simplified) SVM method
- input 2D data and their label (in two classes)
- implement quadratic programming
- output (and plot) classification results
- Programming lanuage: python
- Plot the results in matplotlib
Data and label generated using sklearn:

SVM using linear kernel ('o' for training and 'x' for testing):

SVM using rbf kernel:

SVM using polynomial kernel:

