Selection - Revision history

Carver at 20:29, 23 April 2009

2009-04-23T20:29:51Z

2009-04-23T18:03:07Z

‎Homework

2009-04-23T18:01:09Z

‎Homework

2009-04-23T17:52:05Z

‎An Example Of Overfitting

2009-04-23T16:11:00Z

2009-04-22T23:32:55Z

‎Try it Yourself

2009-04-22T23:31:45Z

‎Try it Yourself

2009-04-22T23:29:17Z

‎Try it Yourself

2009-04-22T23:26:43Z

‎Try it Yourself

2009-04-22T23:12:20Z

‎Try it Yourself

← Older revision		Revision as of 20:29, 23 April 2009
Line 1:		Line 1:
−	~~Today~~ you will fill out teacher evaluations.	+	Next Tuesday you will fill out teacher evaluations.

	Next week will be another good week. Tuesday we will learn to use [[ModelDB\|ModelDB]] and Thursday we will do abrieviated [[Projects\|Projects]].		Next week will be another good week. Tuesday we will learn to use [[ModelDB\|ModelDB]] and Thursday we will do abrieviated [[Projects\|Projects]].

← Older revision		Revision as of 18:03, 23 April 2009
Line 221:		Line 221:
	'''Homework X.1''' For our class of problems when does model selection work well, and when does it work poorly? Try different combinations of low/high levels of noise, and lots of/few data points. For this part, use a fairly simple polynomial to generate the data. Note: I don't know what will happen if you try to fit polynomial with degree higher than the number of data points. Maybe an error.		'''Homework X.1''' For our class of problems when does model selection work well, and when does it work poorly? Try different combinations of low/high levels of noise, and lots of/few data points. For this part, use a fairly simple polynomial to generate the data. Note: I don't know what will happen if you try to fit polynomial with degree higher than the number of data points. Maybe an error.

−	'''Homework X.2''' Investigate the effects of the polynomial's shape on model selection. Use both high and low ~~order~~ polynomials for this problem. Get creative!	+	'''Homework X.2''' Investigate the effects of the polynomial's shape on model selection. Use both high and low degree polynomials for this problem. For low degree polynomials, try, for example, a parabola that is almost a line over the region of the data: (0,1). Get creative!

← Older revision		Revision as of 18:01, 23 April 2009
Line 221:		Line 221:
	'''Homework X.1''' For our class of problems when does model selection work well, and when does it work poorly? Try different combinations of low/high levels of noise, and lots of/few data points. For this part, use a fairly simple polynomial to generate the data. Note: I don't know what will happen if you try to fit polynomial with degree higher than the number of data points. Maybe an error.		'''Homework X.1''' For our class of problems when does model selection work well, and when does it work poorly? Try different combinations of low/high levels of noise, and lots of/few data points. For this part, use a fairly simple polynomial to generate the data. Note: I don't know what will happen if you try to fit polynomial with degree higher than the number of data points. Maybe an error.

−	'''Homework X.2''' Investigate the effects of the polynomial's shape on model selection. Get creative!	+	'''Homework X.2''' Investigate the effects of the polynomial's shape on model selection. Use both high and low order polynomials for this problem. Get creative!

@@ Line 65: / Line 65: @@
 [[Image:Degree2.jpg|center|thumb|300px|Fitting to noisy parabolic data.  Click for full size image]]
-We used a similar parabola to generate the data.  From here on, we start to have too many parameters:
+We used a similar (but undoubtedly not exactly the same) parabola to generate the data.
 Degree 3:

← Older revision		Revision as of 16:11, 23 April 2009
Line 1:		Line 1:
		+	Today you will fill out teacher evaluations.
		+
		+	Next week will be another good week. Tuesday we will learn to use [[ModelDB\|ModelDB]] and Thursday we will do abrieviated [[Projects\|Projects]].
		+
	First, the [[Media:Lab_W.pdf\|homework from Lab W]].		First, the [[Media:Lab_W.pdf\|homework from Lab W]].

@@ Line 207: / Line 207: @@
   LinearData_HiNoise_30pnts = GenPolyData;
-You can also rename data sets (i.e. copy pdata2):
+You can also rename data sets (i.e. copy pdata after it has been defined by GenPolyData and possibly used by the other functions):
-  LinearData_HiNoise_30pnts = pdata2;
+  LinearData_HiNoise_30pnts = pdata;
 == Homework ==

@@ Line 203: / Line 203: @@
   pdata2 = GenPolyData;
-Here I have used a different name (pdata2 instead of pdata) for the structure to save the old data set.  It will only overwrite the data set if you give it the same name.  It might help you to give them more intuitive names:
+Here I have used a different name (pdata2 instead of pdata) for the structure to save the old data set.  It will only overwrite the data set if you give it the same name.  For the homework, which requires you to work with many datasets, might help you to give them more intuitive names:
   LinearData_HiNoise_30pnts = GenPolyData;
 == Homework ==

@@ Line 187: / Line 187: @@
 [[Image:Degree3.jpg|center|thumb|300px|Fitting to noisy parabolic data.  Click for full size image]]
-* The prompt is for a new seed to generate new (cross-validation data).  If you want to start over with a new degree type in -1.
+* The prompt is for a new seed to generate new (cross-validation) data.  If you want to start over with a new degree type in -1.
 * Repeat with new seeds, then type in -1 and repeat with new degrees.  Have fun.

@@ Line 183: / Line 183: @@
   New random seed for Cross-Validation (0 original), or -1 for new degree:
-* Check the figure, it should have plotted the best fitting degree 3 polynomial to the data:
+* It printed the information above and prompted you again.  Also check the figure, it should have plotted the best fitting degree 3 polynomial to the data:
 == Homework ==