next up previous
Next: About this document ... Up: NNA_Exercises_2009 Previous: Boltzmann machines I [3

Boltzmann machines II [2* P]

Train a Boltzmann machine consisting of seven input units and ten output units with the characters of a seven-segment display shown in Figure 2. You have to define the matrix of input patterns and matrix of targets using binary values $ \{-1,+1\}$ according to the Figure.

Figure: Ten numeral patterns for seven-segment display.

a)
Download the file data boltzmann.zip7. Use the Boltzmann machine to classify each of the ten patterns and, thus, verify that most of them have been learned.

b)
Explore the pattern completion in your network in the following way: devise 3-4 deficient patterns with 1 or 2 missing features, which can be unambiguously completed. Use your trained network to learn (complete) these patterns. Try another 3-4 patterns with missing features, but for characters, which can be ambiguously completed.

c)
Add hidden units and compare the performance of the network in completing the ambiguous/unambiguous characters with different number of hidden units.

Present your results clearly, structured and legible.


next up previous
Next: About this document ... Up: NNA_Exercises_2009 Previous: Boltzmann machines I [3
Haeusler Stefan 2010-01-19