Artificial Neural Networks (ANN)

What problem does machine learning solve?

Solving pattern recognition problems

Machine Learning

studies computing algorithms for learning to do stuff.

The emphasis on machine learning is on ______ methods.

automatic

Machine Learning is about learning to do better in the future based on what was experienced in the past.(T/F)

True

Artificial Neural Network

Uses much simpler and abstract neurons, which captures the essence of neural computation. A collection of mathematical functions(instead of collection of neurons).

What are the components of an artificial neural network?

node, input, output, weight, fast

ANN is a brain metaphor for ___________.

information processing

Neural network

A set of connected input/output units where each connection has a weight associated with it

What is ANN good at capturing?

Highly complex non-linear functions

How does the ANN learn in the learning phase?

By adjusting the weights to be able to predict the correct class label of the inputs

What are the uses of ANN?

Prediction(classification, regression) and clustering/segmentation

What are the application areas of ANN?

finance, medicine, marketing, manufacturing, service operations, information systems

What does each node have?

Input and output

What does each node perform?

a simple computation by its node function

_________ connections between nodes.

Weighted

What gives the structure/architecture of the net?

Connectivity

What determines what ANN can compute?

By the connections and their weights

When do you use ANN?

The inputs are well understood, Input is high-dimensional discrete or real valued, Output is discrete or real valued, Output could be a vector of values, Possibly noisy data

What are other things to use ANN for?

Human readability/interpretability of result is unimportant
, Do not have existing knowledge about the patterns of interest, Have a large set of ‘training data’, in data mining and knowledge discovery

What is ANN commonly used for?

Data mining and knowledge discovery

What are the properties of ANN?

Many neuron-like threshold switching units, Highly parallel, distributed process, Emphasis on tuning weights automatically, knows how to process inputs to create an output.
The calculation is quite complex and difficult to understand, yet the results are often useful.

What is the structure of ANN?

A single neuron (processing element – PE) with inputs and outputs

What are the types of activation functions?

Identity, Step, Logistic, TanH, ReLU, Softmax,