Chapter 7: Machine Learning and Deep Learning
1998 Tom Mitchell made a more precise definition for machine learning
A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.
1959 Arthur Samuels coined "machine learning" as
Field of study that gives computers the ability to learn without being explicitly programmed.
deep learning
The most advanced machine learning methods
Recurrent Neural Network (RNN)
a cyclical type of neural network -one to one, one to many, many to one, many to many -h(t-1), h(t), h(t+1)
Kernal
a small matrix used for convolution in image processing -look at small portions of the image separately first
Support Vector Machine
can extrapolate information from one dimensional data (input space) and some information about weights & correlative relationships to another dimension (feature space) -maximizes the width of the decision boundary
machine learning algorithms supervised
continuous -regression, linear, polynomial, decision trees, random forests categorical -classification (KNN, trees, logistic regression, naive babes, SVM)
reinforcement learning
decision process, reward system, learn series of actions
Convolutional Neural Networks (CNN)
deep neural network used for image classification ex:GoogleLeNet
Optimization
deep neural networks are trained through this -where we try to minimize the loss function -
Machine Learning algorithms unsupervised
for continuious -clustering and dimensionality, SVD,PCA, Means for categorical -Association analysis, apriori,FP growth, Hidden Markov model
neural networks
have an input layer, hidden layer and output layer
supervised learning
labeled data, direct feedback, predict outcome/future -develop predictive model based on input and output data -link to classification and regression
unsupervised learning
no labels, no feedback, find hidden structure -group and interpret data based ONLY on input data -links to clustering
logistic regression model
p=1/(1+e^-(b0+b1x)
hyperparameter
parameters used to tune neural network optimization
Deep Learning
refers to a type of machine learning that allows computers to learn complex concepts by learning simpler concepts and combining them -does this using layers that capture simple concepts or build on simpler concepts -a lot of hidden layers
Three types of machine learning
supervised, unsupervised, reinforcement
Machine Learning
the extraction of knowledge from data based on algorithms created from training data -AI learns predictive models extracted from data
recurrent neural networks (RNN)
used for natural language processing and for sequential data
Together RNNs and CNNs can be used for:
•Image captioning •Video classification •Video generation
Convolutional neural networks can be used for:
•Image classification •Video classification •Image captioning •Natural language processing
RNNs can be used for...
•Modelling the stock market •Natural Language Processing (NLP) •Supply chain forecasting •Machine translation •Speech processing
Advanced RNNs
•Recursive neural networks (RNN) •Gated recurrent units (GRUs) •Long short-term memories (LSTMs)
deep neural network (DNN)
Refers to a neural network with more than one hidden layer