MSAA, QXAA, EQAA, and CSAA are post-rendering, while the other AA technologies are post-processing. Post-rendering is more accurate, while Post-processing is lighter and more efficient.
MSAA (Multisampling), aka traditional antialiasing, works by sampling the colors of pixels with adjacent pixels.
QXAA (Quincunx) is one alteration to MSAA, made by Nvidia. It uses the same tech, but reduces load by about 33% by sampling using a Qincunx pattern (a 5-point pattern, only sampling the pixel with adjacent-corner pixels).
CSAA (Coverage sampling) is made by Nvidia, and uses the same tech as MSAA, but allows a higher amount of samples with minimal performance loss by applying extra samples only to pixels that are displaying polygons, on top of an already MSAA-antialiased scene. CSAA uses 4x antialiasing as a base, while a CSAA setting with a 'Q' suffix uses 8x as a base (for example, 16x CSAA uses 4x MSAA with 12x extra coverage samples).
EQAA (Enhanced Quality) is AMD's answer to CSAA. No real difference between the two, other than the different naming scheme. The 'EQ' suffix means that the extra coverage count is equal to the the amount of base samples (4xEQ is 4x MSAA with 4x coverage samples, 8xEQ = 8x MSAA + 8 coverage samples, etc.), while the lack of an 'EQ' suffix denotes an MSAA setting (i.e. no extra coverage samples, EQAA is turned off).
FXAA (Fast approximate) is Nvidia's answer to MLAA, and does exactly what it says: It approximates the pixels to get rid of jaggies as efficiently as possible.
MLAA (Morphological) is a technology made by AMD before FXAA. Exactly the same as FXAA, aside from using a somewhat different algorithm to approximate pixels.
TXAA (Temporal) is a technology made by Nvidia, and uses a combination of a temporal filter with other filters that use information from previous frames to prevent the flickering that you sometimes see with traditional antialiasing.
SMAA (Enhanced Subpixel Morphological) is a technology made by graphics programmers at the University of Zaragoza, using an algorithm that takes some form of MLAA in combination with other technologies like supersampling, TXAA, and sometimes MSAA to render the scene as sharp as possible without being as demanding as MSAA by itself.