toandaominh1997 / cs156-learning-from-data-hw

self-study work for caltech's cs156 machine learning course

learning-from-data-hw

self-study work for caltech's cs156 machine learning course

• hw1: Basic Probabiliy and Perceptron Learning Algorithm DONE
  • perceptron learning algo coded in pla.py
  • problem-specific classes and definitions in hw1_pla.py
  • actual function calling in hw1.py
• hw2: Hoeffding Inequality, Error and Noise, Linear Regression, Nonlinear Transformations DONE
  • testing hoeffding inequality with simulated coin flips in hw2_hoeffding.py
  • error and noise in hw2.txt
  • Linear Regression algo in linreg.py, problem-specific classes/defs in hw2_linreg.py
  • working on nonlinear transformations (hw2_nonlinear.py)
  • actual function calling in hw2.py, call with command-line argument "hoeffding","linreg", or "nlt" (w/o quotes)
• hw3 (hw3.org): Generalization Error, Break Points, Growth Functions, Convex and Non-Convex Sets DONE
• hw4: Generalization Error, Bias and Variance, VC Dimension (hw4.ipynb) DONE
• hw5 (hw5.ipynb): Linear Regression Error, Nonlinear Transforms, Gradient Descent, Logistic Regression, Perceptron Learning Algorithm as Stochastic Gradient Descent DONE
  • logistic regression algo found in logreg.py
• hw6 (hw6.ipynb): Overfitting and Deterministic Noise, Regularization with Weight Decay, Regularization for Polynomials, Neural Networks DONE
  • nonlinear regression algorithm found in hw6_nlt.py
  • course-provided data renamed as hw6_train.dta and hw6_test.dta
  • data loading class found in hw6_dataload.py
• hw7 (hw7.ipynb) DONE: Validation, Validation Bias, Cross Validation, PLA vs SVM
  • PLA coded in pla.py
  • line and line testing algo (with training and testing sets) in line_class.py and line_test.py
  • SVM coded in svm.py
• hw8 (hw8.ipynb) DONE: Primal vs Dual Problem, SVM with Soft Margins: Polynomial Kernels, Cross Validation, RBF Kernel
  • data provided for homework from amlbook.com features.train and features.test and renamed here as hw8_train.dta and hw8_test.dta respectively
  • data loaded and filtered by hw8_dataload.py exporting the LF_Data2 class
  • the SVM class I use comes from scikit-learn
• final