Slow Feature Analysis (SFA) is a method for learning invariant or slowly varying features from vectorial input signals. It involves nonlinear expansion of the input signal and applying principal component analysis to this expanded signal and its time derivative. SFA guarantees to find the optimal solution within a family of functions and can extract a large number of decorrelated features ordered by their degree of invariance. SFA can be applied hierarchically to process high-dimensional input signals and extract complex features. The authors demonstrate SFA's application to complex cell tuning properties based on simple cell outputs, including disparity and motion. They also show how more complicated input-output functions can be approximated by repeated application of SFA. Finally, a hierarchical network of SFA modules is presented as a model of the visual system, capable of learning translation, size, rotation, and contrast invariance for one-dimensional objects. The network's performance degrades when multiple invariances are learned simultaneously.Slow Feature Analysis (SFA) is a method for learning invariant or slowly varying features from vectorial input signals. It involves nonlinear expansion of the input signal and applying principal component analysis to this expanded signal and its time derivative. SFA guarantees to find the optimal solution within a family of functions and can extract a large number of decorrelated features ordered by their degree of invariance. SFA can be applied hierarchically to process high-dimensional input signals and extract complex features. The authors demonstrate SFA's application to complex cell tuning properties based on simple cell outputs, including disparity and motion. They also show how more complicated input-output functions can be approximated by repeated application of SFA. Finally, a hierarchical network of SFA modules is presented as a model of the visual system, capable of learning translation, size, rotation, and contrast invariance for one-dimensional objects. The network's performance degrades when multiple invariances are learned simultaneously.