Python Programming for Biology: Bioinformatics and Beyond

6

In general the grid need not be rectangular. For example, it could be hexagonal.

7

The square brackets are vital so the whole array is ‘multiplied’ by repetition, rather than

the elements inside being numerically multiplied.

8

There are also vstack() and hstack() functions for the other axes.

9

The remainder after integer division, e.g. 13 % 10 = 3 and 3 % 10 = 3.

10

This

is

described

by

the

universal

approximation

theorem;

see

http://http://en.wikipedia.org/wiki/Universal_approximation_theorem

.

11

Although using NumPy can actually be slower for small neural networks.

12

Though it is less obvious what is happening we could also do the following and avoid

the loop: aaIndexDict = dict (enumerate(aminoAcids)).

13

It can, strictly speaking, affect the likelihood of finding a support vector.

14

Aizerman, M., Braverman, E., and Rozonoer, L. (1964). Theoretical foundations of the

potential function method in pattern recognition learning. Automation and Remote Control

25: 821–837.

15

 Boser,  B.E.,  Guyon,  I.M.,  and  Vapnik,  V.N.  (1992).  A  training  algorithm  for  optimal

margin classifiers. In D. Haussler (ed.), 5th Annual ACM Workshop on COLT. Pittsburgh,

PA: ACM Press. pp. 144–152.

16

Since midnight 1 January 1970: the beginning of ‘UNIX time’.

17

Mangasarian, O.L., and Musicant, D.R. (1999). Successive overrelaxation for support

vector machines. IEEE Transactions on Neural Networks 10(5): 1032–1037.

18

Press,  W.H.,  Teukolsky,  S.A.,  Vetterling,  W.T.,  and  Flannery,  B.P.  (2007).  Numerical

---

Recipes.  The  Art  of  Scientific  Computing  (3rd  edn.).  Cambridge:  Cambridge  University

Press.

---