chuckjhao / ai

AI

Basics Of Neural Network

Assigment_1

• Build a neural network that predicts the price of a house according to a simple formula.

Assignment_mnist

• Handwriting digits 0 through 9, Use your own image.

BasicNN

HorseOrHumanLesson 3 - Notebook

• This code will allow you to choose 1 or more files from your file system, it will then upload them, and run them through the model, giving an indication of whether the object is a horse or a human.

RNNBasic

SimpleConvNet

• Improving Computer Vision Accuracy using Convolutions, Visualizing the Convolutions and Pooling.

fashionMNISTNotebook

---

---

Machine learning

Common

• init, utility.

Data Preprocess

• Similarity

  • Article classification compare similarity using angle similarity.

• preprocess

  • Encoding Ordinal features, Encoding Categorical features, one hot encoding, Check for missing values, Missing value processing, Fill in missing values with mean, outlier.

  • Project: Iris, Train_test_split, Standardization (z-score), Normaliaztion, MinMaxScaler, Normalization.

KNN

• Preparing the Data, Preprocessing, Train Test Split, Feature Scaling, Training and Predictions, Evaluating the Algorithm, Comparing Error Rate with the K Value.

K-means

• Genearte sample data, K-means algorithm, Plot Scatter, Number of iterations run, Make a circle Dataset, Manually set centroid, Apply K-means with re-scaled data.

Logistic Regression

• LogisticRegression

  • Import Iris data set, Feature scaling, Standardization, Train, Calculation verification, Visualize training data classification results, Calculate the number of misclassified test data, Output prediction probability.

• Logistic_multiclass decision regions

  • Import Iris data set, decision_regions for test data, getting the confusion matrix, seaborn pairplot.

• Logistic_multiclass

  • Import Iris data set, decision_regions for test data, decision_regions for training data, getting the confusion matrix.

Decision Tree and Random Forest

• Decision Tree

  • Train, Classifier, Use entropy as a criterion, Calculation accuracy, Test report, Decision tree visualization, Visualize the decision boundary of the decision tree.

• Random Forest

  • Calculate the score, Use RandomForest to find out the main features of Iris data classification, AdaBoost (Adaptive Boosting) Algorithm.

Regression

• Boston_House_Price

  • Project: House-Price-Prediction use Linear Regression, Basic data analysis, train test split, Model fit, Prediction, Evaluation model, Save/Export Model, Plot, Differences with or without standardization, k-fold cross-validation : evaluating estimator performance.

• Linear Regression-1

  • Linear, Nonlinear, Training data, Test data, Calculate MSE.

• Linear Regression-2

  • Property value prediction, Boston House Price, cross-validation.

• Ridge and Lasso Regression

  • Create a Ridge Regression, R2 Score, Create a Lasso Regression.

---

---

Deep learning

Concept

• activation

  • sigmod, Relu, Tangent, Softmax, Cross Entropy, ACE(Average Cross Entropy).

• exercise1

  • Create a logical gate using a simple DNN.

• hello_keras

  • Define Network, Prepare data, training, model evaluation, model prediction, model score.

MLP from scratch

• DeepLearning without framework numeric method

  • Use numerical differentiation method to find the differential (partial differential) solution of the function, gradient, Plot a 2D field of arrows, Define two layers network, load MNIST dataset, create the model, train the model, save the model, Load pre-trained model.

Deep Neural Network

• ANN_regression_Boston_House_Price

  • Project: House-Price-Prediction - use Neural Network, Objectives:1. Predict the sale price for each house. 2. Minimize the difference between predicted and actual rating (RMSE/MSE).

• DNN,AND,OR,XOR Gate

• iris_DNN

  • Project: Iris, Load DataSet, Data Preproecssing, Create a NN model using Keras, Training, Evaluation, Make predictions, Plot scatter matrix, Plot confusion matrix.

Recurrent Neural Network

• airline-passenger

  • Project: RNN_passengers, Loading and Visualizing Data, Preprocessing Data, Create SimpleRNN Model, Predictions and Visualising RNN Model.

• KKTV

  • Project: train.

• RNN exercise2

  • Project: Create an RNN network that can predict the next number, Generate the input sequence, create model and train model, Test.

• RNN exercise2 power

  • Project: Create an RNN network that can predict the next number, Generate the input sequence, create model and train model, Test.

Convolutional Neural Network

• convolution

  • padding='same', Import a picture for convolution and output the picture after convolution (street, test, window), Edge Detection, Blur.

---

GitHub @ChuckJhao