Machine Learning
How can we get machines to learn from observations?
We want a machine to recognize write digits, also with low pixels
Different kinds of tasks:
Classification:
Assigning data to a class;
- Linear
- Non-linear
- Complex
- Multiple categories
Signal classification:
Forecasting = learning from observations; getting some rules that can be
used in the future to predict the next outcomes (for example in finance).
Predicting the next outcomes of a time series
Function approximation = determining the value of a point
- Identify species (such as birds or insects)
- Identify abnormalities (such as irregular heart rate)
Unsupervised clustering:
Find unknown clusters in data
- Species assemblages
- Protein structure
- Clustering your customers
Learning from observations:
Example 1 – predicting the water
- Regression problem output is real/continuous (salary/weight)
- Learning to estimate a numeric output predicting the tempearture
- Problem of overfitting applies to every machine learning problem
Observation = series of k observations (examples/instances/cases), and observation
describes a set of inputs x (x1,..,xn) and an output y. Each xi is called a
feature/attribute/input variable. Y is typically called the output variable.
There is a related input and output, the input is a single value.
Feature is the day of the year we are talking about
How accurately can we predict the temperature? accurate prediction is hard
How can we get machines to learn from observations?
We want a machine to recognize write digits, also with low pixels
Different kinds of tasks:
Classification:
Assigning data to a class;
- Linear
- Non-linear
- Complex
- Multiple categories
Signal classification:
Forecasting = learning from observations; getting some rules that can be
used in the future to predict the next outcomes (for example in finance).
Predicting the next outcomes of a time series
Function approximation = determining the value of a point
- Identify species (such as birds or insects)
- Identify abnormalities (such as irregular heart rate)
Unsupervised clustering:
Find unknown clusters in data
- Species assemblages
- Protein structure
- Clustering your customers
Learning from observations:
Example 1 – predicting the water
- Regression problem output is real/continuous (salary/weight)
- Learning to estimate a numeric output predicting the tempearture
- Problem of overfitting applies to every machine learning problem
Observation = series of k observations (examples/instances/cases), and observation
describes a set of inputs x (x1,..,xn) and an output y. Each xi is called a
feature/attribute/input variable. Y is typically called the output variable.
There is a related input and output, the input is a single value.
Feature is the day of the year we are talking about
How accurately can we predict the temperature? accurate prediction is hard
