<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=ISO-8859-1">
</head>
<body bgcolor="#FFFFFF" text="#000000">
Hello class<br>
<br>
Here are a couple of clarifications about homework 5:<br>
* In 1.b), "hard-coded" means you can write a one-line functions
which just return a vector "return np.array([1,5,...])". The
functions don't have to contain any other code. Of course, you have
to find a way to get to a good solution (by hand is fine, but check
that your solution does obey the requirement).<br>
<br>
* Uniform allocation means uniform number of bits <b>per line, not
per band.</b><br>
<br>
* In the uniform allocation, you can allocate the same number of
bits per line in each band but it's also ok to allocate one more bit
to some bands if you've got more to allocate. Here "uniform" would
mean that all the bands have to have <u>either n or n+1 bit/line</u>.
Know that for coder, lower bands are usually privileged for the
attribution of extra bits.<br>
<br>
* <u>Every function has to return the number of bits allocated <b>by
line</b> in each band</u>. For example if the function <u>returns
an array whose first element is 4</u>, that means that for the
band n°1, <u>you'll use 4*nLines[0]</u> (index 0 for first band)
from the budget.<br>
<br>
* A new version of bitalloc.pyc has been uploaded on the website.<br>
<br>
* In the scaleFactorBands class, be careful that the convention is
that <u>to access all the lines of a band</u>, you call
MDCTdata[lowerLine[i]:upperLine[i]<b><u>+1</u></b>].<br>
<br>
<br>
Please let me know if any of this is unclear,<br>
Best,<br>
<pre class="moz-signature" cols="72">--
François Germain
Teaching Assistant / PhD Candidate
Center for Computer Research in Music and Acoustics (CCRMA)
Stanford University</pre>
</body>
</html>