6.034 Fall 2009 - Recent changes [en]

Tutorials

Yks — Tue, 24 Nov 2009 03:49:23 GMT

Changing Radhika to Yuan's Monday 3-4 section

←Older revision		Revision as of 03:49, 24 November 2009
(2 intermediate revisions not shown.)
Line 268:		Line 268:
	\|-		\|-
	\| pfuller \|\| Pierre Henri Fuller \|\| Yuan K. Shen \|\| Monday 3--4pm \|\| 38-166		\| pfuller \|\| Pierre Henri Fuller \|\| Yuan K. Shen \|\| Monday 3--4pm \|\| 38-166
		+	\|-
		+	\| radhikam \|\| Radhika Malik \|\| Yuan K. Shen \|\| Monday 3--4pm \|\| 38-166
	\|-		\|-
	\| \|\| \|\| \|\| \|\|		\| \|\| \|\| \|\| \|\|
Line 374:		Line 376:
	\|-		\|-
	\| vwang11 \|\| Victoria Wang \|\| Matthew Peairs \|\| Tuesday 1--2pm \|\| 34-302		\| vwang11 \|\| Victoria Wang \|\| Matthew Peairs \|\| Tuesday 1--2pm \|\| 34-302
-	\|-
-	~~\| radhikam \|\| Radhika Malik \|\| Matthew Peairs \|\| Tuesday 1--2pm \|\| 34-302~~
	\|-		\|-
	\| \|\| \|\| \|\| \|\|		\| \|\| \|\| \|\| \|\|

Lab 5

Yks — Mon, 23 Nov 2009 22:06:19 GMT

←Older revision		Revision as of 22:06, 23 November 2009
(16 intermediate revisions not shown.)
Line 8:		Line 8:
	* Or, on Athena, <tt>add 6.034</tt> and copy it from <tt>/mit/6.034/www/labs/lab5/</tt>.		* Or, on Athena, <tt>add 6.034</tt> and copy it from <tt>/mit/6.034/www/labs/lab5/</tt>.
	* <font size="+1">You will need to <font size="+2"><b>download and install</b> an additional software package called [http://www.ailab.si/orange/ Orange]</font>. Do this first so that you get the problems worked out early. If you have downloaded and installed it, you should be able to run lab5.py with no changes, and see the output of several classifiers on the vampire dataset.</font> If you get errors, email us.		* <font size="+1">You will need to <font size="+2"><b>download and install</b> an additional software package called [http://www.ailab.si/orange/ Orange]</font>. Do this first so that you get the problems worked out early. If you have downloaded and installed it, you should be able to run lab5.py with no changes, and see the output of several classifiers on the vampire dataset.</font> If you get errors, email us.
-	* Note: Orange is available for Linux, Windows, and OS X (and we have ~~successful,~~ tested ~~psets~~ on ~~those~~ platforms), ~~but Athena needs a newer version of quicktime for it to work~~. <b>If you ~~need to~~ work on Athena</b>~~, let us~~ know, ~~and~~ we'~~ll speed our efforts at~~ a ~~solution~~.	+	* Note: Orange is available for Linux (Ubuntu), Windows, and OS X (and we have successfully tested this lab on these platforms). Also, if you are working on Mac OS X, be sure that you have [http://sourceforge.net/projects/numpy/files/NumPy/1.3.0/numpy-1.3.0-py2.6-macosx10.5.dmg/download Numpy] installed.
		+	* To check that your Orange is properly installed, run: <tt>python orange_for_6034.py</tt> and you should get a version string and no errors.
		+
		+	To work on this lab on Athena you'll need to:
		+	# If you use bash: Setup your environment to pickup the class installation of Orange by adding the following: <pre>export LD_LIBRARY_PATH=/afs/athena.mit.edu/course/6/6.034/lib/python2.6/dist-packages/orange:$LD_LIBRARY_PATH</pre> to your .bashrc.mine file. If you are using tcsh or csh run: <pre>setenv LD_LIBRARY_PATH /afs/athena.mit.edu/course/6/6.034/lib/python2.6/dist-packages/orange:$LD_LIBRARY_PATH</pre> instead. You may omit <tt>:$LD_LIBRARY_PATH</tt> if you get complaints about it not being set. The above settings tell python where to look for shared libraries required for running Orange. NOTE: If you don't know what the shell you are running, type "echo $SHELL".
		+	# Log into linux.mit.edu: <tt>ssh -X <user>@linux.mit.edu</tt>
		+	# For your convenience, we've provided a script, <tt>run-orange-gui.sh</tt> that will run the Orange GUI on linux.mit.edu. Note that some Orange Widgets (like ROC charting) may not appear in GUI in the Athena version. Don't worry, you won't need the GUI to answer the questions for this lab.

	Your answers for the problem set belong in the main file <tt>lab5.py</tt>, as well as <tt>boost.py</tt>.		Your answers for the problem set belong in the main file <tt>lab5.py</tt>, as well as <tt>boost.py</tt>.
Line 59:		Line 65:
	=== Using Orange from Python ===		=== Using Orange from Python ===

-	We have given you a function called describe_and_classify that trains a bunch of classifiers that Orange provides. Some of them will be more familiar than others.	+	We have given you a function called <tt>describe_and_classify</tt> that trains a bunch of classifiers that Orange provides. Some of them will be more familiar than others.

	First it trains each classifier on the data, shows its output on each data point from the training data, and shows the accuracy on the training data. You know from class that the accuracy on the training data should be 1 for these classifiers. It is less than one because each classifier comes with built-in regularization to help prevent overtraining. That's what the pruning is for the decision tree, for example. We didn't specify regularization parameters to most of the learners because they have fine defaults. You can read more about them in the Orange documentation.		First it trains each classifier on the data, shows its output on each data point from the training data, and shows the accuracy on the training data. You know from class that the accuracy on the training data should be 1 for these classifiers. It is less than one because each classifier comes with built-in regularization to help prevent overtraining. That's what the pruning is for the decision tree, for example. We didn't specify regularization parameters to most of the learners because they have fine defaults. You can read more about them in the Orange documentation.
Line 65:		Line 71:
	You'll notice that we do one extra thing with the decision tree. We print it out. For most classifiers, their internal representations are hard for humans to read, but the decision tree's internal representation can be very useful in getting to know your data.		You'll notice that we do one extra thing with the decision tree. We print it out. For most classifiers, their internal representations are hard for humans to read, but the decision tree's internal representation can be very useful in getting to know your data.

-	Then describe_and_classify passes the untrained learners, without having seen the data, to cross-validation. Cross-validation hides part of the training data, trains the learner on the rest, and then tests it on the hidden part. To avoid accidentally picking just one terrible subset of the data to hide (an irreproducible subset), it divides the data evenly into some number of folds, successively tests by hiding each fold in turn, and averages over the results. You will find with cross-validation that the classifiers do much better on identifying political parties than they do on vampires. That's because the vampire dataset has so few examples, with so little redundancy, that if you take away one example, it is very likely to remove the information you actually need in order to classify that example.	+	Then <tt>describe_and_classify</tt> passes the untrained learners, without having seen the data, to cross-validation. Cross-validation hides part of the training data, trains the learner on the rest, and then tests it on the hidden part. To avoid accidentally picking just one terrible subset of the data to hide (an irreproducible subset), it divides the data evenly into some number of folds, successively tests by hiding each fold in turn, and averages over the results. You will find with cross-validation that the classifiers do much better on identifying political parties than they do on vampires. That's because the vampire dataset has so few examples, with so little redundancy, that if you take away one example, it is very likely to remove the information you actually need in order to classify that example.

-	To ensure that you've gotten the right idea from each part of describe_and_classify, there are <b>six questions</b> just below the function.	+	To ensure that you've gotten the right idea from each part of <tt>describe_and_classify</tt>, there are <b>six questions</b> just below the function.

	=== Boosting with Orange ===		=== Boosting with Orange ===

Lab 5

Gremio — Thu, 19 Nov 2009 13:53:35 GMT

d&c doesn't actually do the creating anymore

←Older revision		Revision as of 13:53, 19 November 2009
(2 intermediate revisions not shown.)
Line 14:		Line 14:
	== Boosting ==		== Boosting ==

-	You're still trying to use AI to predict the votes of politicians. ~~You decide that the~~ ID-~~tree classifier was too rigid~~ and ~~uninformative~~, so ~~now~~ you ~~try using a boosting classifier instead~~.	+	You're still trying to use AI to predict the votes of politicians. ID-Trees were great, but you've heard about these other magnificent learning algorithms like SVMs and Boosting. Boosting sounds easier to implement and had a pretty good reputation, so you decide to start with that.

	To make sure that you interpret the results without letting your political preconceptions get in the way, you dig up some old data to work with: in particular, the data from the 4th House of Representatives, which met from 1796 to 1797. (According to the records on [http://www.voteview.com voteview.com], this is when the two-party system first emerged, with the two parties being designated "Federalists" and "Republicans".)		To make sure that you interpret the results without letting your political preconceptions get in the way, you dig up some old data to work with: in particular, the data from the 4th House of Representatives, which met from 1796 to 1797. (According to the records on [http://www.voteview.com voteview.com], this is when the two-party system first emerged, with the two parties being designated "Federalists" and "Republicans".)
Line 59:		Line 59:
	=== Using Orange from Python ===		=== Using Orange from Python ===

-	We have given you a function called describe_and_classify that ~~creates and~~ trains a bunch of classifiers that Orange provides. Some of them will be more familiar than others.	+	We have given you a function called describe_and_classify that trains a bunch of classifiers that Orange provides. Some of them will be more familiar than others.

	First it trains each classifier on the data, shows its output on each data point from the training data, and shows the accuracy on the training data. You know from class that the accuracy on the training data should be 1 for these classifiers. It is less than one because each classifier comes with built-in regularization to help prevent overtraining. That's what the pruning is for the decision tree, for example. We didn't specify regularization parameters to most of the learners because they have fine defaults. You can read more about them in the Orange documentation.		First it trains each classifier on the data, shows its output on each data point from the training data, and shows the accuracy on the training data. You know from class that the accuracy on the training data should be 1 for these classifiers. It is less than one because each classifier comes with built-in regularization to help prevent overtraining. That's what the pruning is for the decision tree, for example. We didn't specify regularization parameters to most of the learners because they have fine defaults. You can read more about them in the Orange documentation.
Line 65:		Line 65:
	You'll notice that we do one extra thing with the decision tree. We print it out. For most classifiers, their internal representations are hard for humans to read, but the decision tree's internal representation can be very useful in getting to know your data.		You'll notice that we do one extra thing with the decision tree. We print it out. For most classifiers, their internal representations are hard for humans to read, but the decision tree's internal representation can be very useful in getting to know your data.

-	Then describe_and_classify ~~creates~~ the ~~same~~ learners, ~~but~~ without the data, ~~and passes them~~ to cross-validation. Cross-validation hides part of the training data, trains the learner on the rest, and then tests it on the hidden part. To avoid accidentally picking just one terrible subset of the data to hide (an irreproducible subset), it divides the data evenly into some number of folds, successively tests by hiding each fold in turn, and averages over the results. You will find with cross-validation that the classifiers do much better on identifying political parties than they do on vampires. That's because the vampire dataset has so few examples, with so little redundancy, that if you take away one example, it is very likely to remove the information you actually need in order to classify that example.	+	Then describe_and_classify passes the untrained learners, without having seen the data, to cross-validation. Cross-validation hides part of the training data, trains the learner on the rest, and then tests it on the hidden part. To avoid accidentally picking just one terrible subset of the data to hide (an irreproducible subset), it divides the data evenly into some number of folds, successively tests by hiding each fold in turn, and averages over the results. You will find with cross-validation that the classifiers do much better on identifying political parties than they do on vampires. That's because the vampire dataset has so few examples, with so little redundancy, that if you take away one example, it is very likely to remove the information you actually need in order to classify that example.

	To ensure that you've gotten the right idea from each part of describe_and_classify, there are <b>six questions</b> just below the function.		To ensure that you've gotten the right idea from each part of describe_and_classify, there are <b>six questions</b> just below the function.

Labs

Gremio — Thu, 19 Nov 2009 05:00:33 GMT

Currently released labs:

←Older revision		Revision as of 05:00, 19 November 2009
Line 6:		Line 6:
	* [[Lab 3]] -- due Thursday, October 15		* [[Lab 3]] -- due Thursday, October 15
	* [[Lab 4]] -- due Friday, October 30		* [[Lab 4]] -- due Friday, October 30
		+	* [[Lab 5]] -- due Friday, December 4

	== The online grader ==		== The online grader ==

Lab 5

Gremio — Thu, 19 Nov 2009 04:00:26 GMT

added bonus

←Older revision		Revision as of 04:00, 19 November 2009
(15 intermediate revisions not shown.)
Line 7:		Line 7:
	* Download it as a ZIP file: http://web.mit.edu/6.034/www/labs/lab5/lab5.zip		* Download it as a ZIP file: http://web.mit.edu/6.034/www/labs/lab5/lab5.zip
	* Or, on Athena, <tt>add 6.034</tt> and copy it from <tt>/mit/6.034/www/labs/lab5/</tt>.		* Or, on Athena, <tt>add 6.034</tt> and copy it from <tt>/mit/6.034/www/labs/lab5/</tt>.
		+	* <font size="+1">You will need to <font size="+2"><b>download and install</b> an additional software package called [http://www.ailab.si/orange/ Orange]</font>. Do this first so that you get the problems worked out early. If you have downloaded and installed it, you should be able to run lab5.py with no changes, and see the output of several classifiers on the vampire dataset.</font> If you get errors, email us.
		+	* Note: Orange is available for Linux, Windows, and OS X (and we have successful, tested psets on those platforms), but Athena needs a newer version of quicktime for it to work. <b>If you need to work on Athena</b>, let us know, and we'll speed our efforts at a solution.

	Your answers for the problem set belong in the main file <tt>lab5.py</tt>, as well as <tt>boost.py</tt>.		Your answers for the problem set belong in the main file <tt>lab5.py</tt>, as well as <tt>boost.py</tt>.
-
-	~~==Neural Nets==~~
-	In this problem set, you will be experimenting with neural nets. Although you'll only have to do the most superficial of coding, take your time and make sure you really understand what is happening when you run the nets.
-
-	~~===The training process===~~
-	In this problem set, we'll ask you to train a neural net until it converges. (We've written a function that does this for you; see the <tt>lab5.py</tt> file.) Here's how we define convergence for the purposes of this problem set:
-	* Train the neural net in increments of 100 epochs. (An epoch is a run through the entire set of training data.)
-	* After each 100 epochs, check to see if the neural net has ''converged'', which for the purpose of this problem set we define as follows:
-	** The mean error of the neural net has changed by less than .0001 since 100 epochs ago.
-	** The mean error has changed by more than .0001 in ''some'' previous group of 100 epochs. In other words, a neural net hasn't converged yet if it's still trying to get out of its start state.
-	* Stop training when the neural net has converged or 200,000 epochs have passed, whichever happens first.
-
-	~~These neural nets are initialized with small random weights, so it is expected that they will have varying behavior.~~
-
-	~~===Learning XOR===~~
-	~~The first function the neural net will be learning is XOR. XOR is a binary function that returns a 1 when exactly one of its two inputs is 1. The XOR table looks like this:~~
-
-	~~A B XOR(A, B)~~
-	~~1 1 0~~
-	~~1 0 1~~
-	~~0 1 1~~
-	~~0 0 0~~
-
-	~~To do this, we are using a net with 2 input neurons, 2 hidden (internal) neurons, and 1 output neuron. Take a look near the top to find the procedures you will use in training the XOR function.~~
-
-	Training the neural net is achieved by calling <tt>train_neural_net(neural_network, examples, learning_rate)</tt>. This trains the neural net on the XOR data for up to 200,000 epochs (or until it converges), and returns the average error of the neural net when it is done.
-
-	#Initialize the XOR net with a learning rate of 0.5, and train it until it converges. Do this a few times to make sure that your results are consistent. Report the average error at the end of training, and the average number of epochs you trained for, in <tt>lab5.py</tt>.
-	~~#Do the same with a learning rate of 0.05, and report your results in <tt>lab5.py</tt>.~~
-	~~#Try several different learning rates. What learning rate takes the fewest epochs, on average, to converge?~~
-	~~#Change the number of internal nodes to 4 (With learning rate set at 0.5) How many epochs on average does it take to converge?~~
-	~~#Change the number of internal nodes to 6. How many epochs on average does it take to converge?~~
-	~~#What is the minimum number of internal nodes you need for the neural net to learn this function with better than 99% accuracy?~~
-
-	~~=== Checking your answers ===~~
-	You can sanity-check your answers with <tt>tester.py</tt>. Don't worry about being "exactly right"; there will be a wide range of acceptable answers. If you are confident of your answer to within 20%, you are fine.

	== Boosting ==		== Boosting ==
Line 73:		Line 39:
	Here are the parts that you need to complete:		Here are the parts that you need to complete:
	* In the <tt>BoostClassifier</tt> class in <tt>boost.py</tt>, the <tt>update_weights</tt> method is undefined. You need to define this method so that it changes the data weights in the way prescribed by the AdaBoost algorithm. There are two ways of implementing this update which happen to be mathematically equivalent.		* In the <tt>BoostClassifier</tt> class in <tt>boost.py</tt>, the <tt>update_weights</tt> method is undefined. You need to define this method so that it changes the data weights in the way prescribed by the AdaBoost algorithm. There are two ways of implementing this update which happen to be mathematically equivalent.
-	* In the <tt>BoostClassifier</tt> class, the <tt>classify</tt> method is also undefined. Define it so that you can use a trained BoostClassifier as a classifier, outputting +1 or -1 based on the weighted results of its base classifiers.	+	* In the <tt>BoostClassifier</tt> class, the <tt>classify</tt> method is also undefined. Define it so that you can use a trained BoostClassifier as a classifier, outputting +1 or -1 based on the weighted results of its base classifiers. Complete the very similar <tt>orange_classify</tt> method as well.
	* In <tt>lab5.py</tt>, the <tt>most_misclassified</tt> function is undefined. You will need to define it to answer the questions.		* In <tt>lab5.py</tt>, the <tt>most_misclassified</tt> function is undefined. You will need to define it to answer the questions.

	=== Questions ===		=== Questions ===
-	Answer the questions in <tt>lab5.py</tt>.	+	Answer the two questions <tt>republican_newspaper_vote</tt> and <tt>republican_sunset_vote</tt> in <tt>lab5.py</tt>.

	When you are asked how a particular political party would vote on a particular motion, disregard the possibility of abstaining. If your classifier results indicate that the party ''wouldn't'' vote NO, consider that an indication that the party would vote YES.		When you are asked how a particular political party would vote on a particular motion, disregard the possibility of abstaining. If your classifier results indicate that the party ''wouldn't'' vote NO, consider that an indication that the party would vote YES.
		+
		+	== Orange you glad someone else implemented these? ==
		+
		+	First things first: Have you installed Orange yet?
		+
		+	Now that you've installed Orange, when you run lab5.py, does it complain about Orange, or does it show you the outputs of some classifiers on vampire data?
		+
		+	=== Getting familiar with Orange ===
		+
		+	This part is optional: it's about using the Orange GUI to do a little machine learning without doing any programming. We've given you some data files (vampires.tab, H004.tab, adult.tab, titanic.tab, breast-cancer.tab, etc.) that you can play with. Try making something like [http://people.csail.mit.edu/gremio/tmp/6.034/orange-screenshot.png this screenshot], and look at the performance, and look at the actual predictions.
		+
		+	=== Using Orange from Python ===
		+
		+	We have given you a function called describe_and_classify that creates and trains a bunch of classifiers that Orange provides. Some of them will be more familiar than others.
		+
		+	First it trains each classifier on the data, shows its output on each data point from the training data, and shows the accuracy on the training data. You know from class that the accuracy on the training data should be 1 for these classifiers. It is less than one because each classifier comes with built-in regularization to help prevent overtraining. That's what the pruning is for the decision tree, for example. We didn't specify regularization parameters to most of the learners because they have fine defaults. You can read more about them in the Orange documentation.
		+
		+	You'll notice that we do one extra thing with the decision tree. We print it out. For most classifiers, their internal representations are hard for humans to read, but the decision tree's internal representation can be very useful in getting to know your data.
		+
		+	Then describe_and_classify creates the same learners, but without the data, and passes them to cross-validation. Cross-validation hides part of the training data, trains the learner on the rest, and then tests it on the hidden part. To avoid accidentally picking just one terrible subset of the data to hide (an irreproducible subset), it divides the data evenly into some number of folds, successively tests by hiding each fold in turn, and averages over the results. You will find with cross-validation that the classifiers do much better on identifying political parties than they do on vampires. That's because the vampire dataset has so few examples, with so little redundancy, that if you take away one example, it is very likely to remove the information you actually need in order to classify that example.
		+
		+	To ensure that you've gotten the right idea from each part of describe_and_classify, there are <b>six questions</b> just below the function.
		+
		+	=== Boosting with Orange ===
		+
		+	You may be able to do better by using AdaBoost over your Orange classifiers than you could do with any of those Orange classifiers by themselves. Then again, you may do worse. That AdaBoost "doesn't tend to overfit" is a somewhat conditional statement that depends a lot on the particular classifiers that boosting gets its pick of. If some of the underlying classifiers tend to overfit, then boosting will greedily choose them first, and it will be stuck with those choices for testing.
		+
		+	We have set up a learner that uses the BoostClassifier from the first part, but its underlying classifiers are the Orange classifiers we were just looking at. When you combine classifiers in this way, you create what's called an "ensemble classifier". You will notice, as you run this new classifier on the various data sets we've provided, that the ensemble frequently performs worse in cross-validation than some (or most) of its underlying classifiers.
		+
		+	Your job is to find a set of classifiers for the ensemble that get <b>at least 74% accuracy on the breast-cancer dataset</b>. You may use any subset of the classifiers we've provided, and you may use the same classifier more than once. Put the short names of your classifiers into the list <tt>classifiers_for_best_ensemble</tt>. There will be honorable mention in class for the best ensemble. If you are proud of the way that you went about choosing the best ensemble, let us know to look at your code carefully, and there may be another honorable mention for that.
		+
		+	=== Bonus! ===
		+
		+	We think there may be a problem with the code somewhere, but we don't know quite where. If you run the ensemble boosting with the original set of learners (one of each) then you get terrible results for the H109 data set. Can you figure out why?

	== Errata ==		== Errata ==
-	Results were varying on <tt>most_misclassified</tt>. The cause seemed to be that floating-point error was prioritizing different classifiers for different people, causing different data points to be emphasized despite that the results were about the same. In response, we've replaced the <tt>most_misclassified</tt> test with a ~~very weak sanity~~ check.	+	We expect a few: check back.

Lab 5

Yks — Tue, 17 Nov 2009 21:48:58 GMT

←Older revision		Revision as of 21:48, 17 November 2009
(One intermediate revision not shown.)
Line 1:		Line 1:
	__TOC__		__TOC__

-	This is the last problem set in 6.034! It's due Friday, December ~~5th~~ at 11:59 PM.	+	This is the last problem set in 6.034! It's due Friday, December 4th at 11:59 PM.

	To work on this problem set, you will need to get the code:		To work on this problem set, you will need to get the code:
Line 39:		Line 39:
	#Do the same with a learning rate of 0.05, and report your results in <tt>lab5.py</tt>.		#Do the same with a learning rate of 0.05, and report your results in <tt>lab5.py</tt>.
	#Try several different learning rates. What learning rate takes the fewest epochs, on average, to converge?		#Try several different learning rates. What learning rate takes the fewest epochs, on average, to converge?
-	#Change the number of internal nodes to 4. How many epochs on average does it take to converge?	+	#Change the number of internal nodes to 4 (With learning rate set at 0.5) How many epochs on average does it take to converge?
	#Change the number of internal nodes to 6. How many epochs on average does it take to converge?		#Change the number of internal nodes to 6. How many epochs on average does it take to converge?
	#What is the minimum number of internal nodes you need for the neural net to learn this function with better than 99% accuracy?		#What is the minimum number of internal nodes you need for the neural net to learn this function with better than 99% accuracy?
Line 58:		Line 58:
	The following resources will be helpful:		The following resources will be helpful:
	* The documentation for the boosting code, which you can find embedded in <tt>boost.py</tt> in the documentation strings.		* The documentation for the boosting code, which you can find embedded in <tt>boost.py</tt> in the documentation strings.
-	* The [http://courses.csail.mit.edu/6.034f/ai3/msri.pdf ~~course notes~~ on boosting], or the [http://~~web~~.mit.edu/6.~~034~~/~~psets-f07~~/~~ps5/boost_notes~~.pdf ~~handout~~] that summarizes it.	+	* The [http://courses.csail.mit.edu/6.034f/ai3/msri.pdf Shapire paper on boosting], or the [http://courses.csail.mit.edu/6.034f/ai3/boosting.pdf notes] that summarizes it.
	* The [[Lab 4]] writeup, if you need to refer to how <tt>data_reader.py</tt> represents legislators and votes.		* The [[Lab 4]] writeup, if you need to refer to how <tt>data_reader.py</tt> represents legislators and votes.

Line 82:		Line 82:

	== Errata ==		== Errata ==
-	The original version of the lab had the <tt>BoostClassifier.reset()</tt> method in <tt>boost.py</tt> defined incorrectly. Where it says <tt>data</tt>, it should be <tt>self.data</tt>. A fixed version is uploaded.
-
-	~~A comment in the neural net section is wrong. You should give us the average number of ''epochs'' with 4 or 6 hidden neurons, as the variable names say, not the average error like the comment says.~~
-
	Results were varying on <tt>most_misclassified</tt>. The cause seemed to be that floating-point error was prioritizing different classifiers for different people, causing different data points to be emphasized despite that the results were about the same. In response, we've replaced the <tt>most_misclassified</tt> test with a very weak sanity check.		Results were varying on <tt>most_misclassified</tt>. The cause seemed to be that floating-point error was prioritizing different classifiers for different people, causing different data points to be emphasized despite that the results were about the same. In response, we've replaced the <tt>most_misclassified</tt> test with a very weak sanity check.

Index

Yks — Tue, 17 Nov 2009 21:39:42 GMT

←Older revision		Revision as of 21:39, 17 November 2009
(3 intermediate revisions not shown.)
Line 15:		Line 15:
	* [http://courses.csail.mit.edu/6.034f/Examinations/2008q1.pdf 2008 Quiz 1 P1 (Dinosaurs)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s1.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2008q1.pdf 2008 Quiz 1 P1 (Dinosaurs)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s1.pdf solution]]
	* [http://courses.csail.mit.edu/6.034f/Examinations/2009q1.pdf 2009 Quiz 1 P1 (Harry Potter)] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s1.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2009q1.pdf 2009 Quiz 1 P1 (Harry Potter)] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s1.pdf solution]]
		+

	Search		Search
Line 35:		Line 36:
	* [http://courses.csail.mit.edu/6.034f/Examinations/2008q1.pdf 2008 Quiz 1 P2 (Depth-First-Search Star)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s1.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2008q1.pdf 2008 Quiz 1 P2 (Depth-First-Search Star)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s1.pdf solution]]
	* [http://courses.csail.mit.edu/6.034f/Examinations/2009q1.pdf 2009 Quiz 1 P2 (Subway)] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s1.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2009q1.pdf 2009 Quiz 1 P2 (Subway)] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s1.pdf solution]]
		+

	Games		Games
Line 47:		Line 49:
	* [http://courses.csail.mit.edu/6.034f/Examinations/2008q2.pdf 2008 Quiz 2 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s2.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2008q2.pdf 2008 Quiz 2 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s2.pdf solution]]
	* [http://courses.csail.mit.edu/6.034f/Examinations/2009q2.pdf 2009 Quiz 2 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s2.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2009q2.pdf 2009 Quiz 2 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s2.pdf solution]]
		+

	Constraint Satisfaction Problems (CSP)		Constraint Satisfaction Problems (CSP)
Line 67:		Line 70:
	* [http://courses.csail.mit.edu/6.034f/Examinations/2008q2.pdf 2008 Quiz 2 P2 (Seating Moose and Friends)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s2.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2008q2.pdf 2008 Quiz 2 P2 (Seating Moose and Friends)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s2.pdf solution]]
	* [http://courses.csail.mit.edu/6.034f/Examinations/2009q2.pdf 2009 Quiz 2 P2 (Time traveling lecturers)] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s2.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2009q2.pdf 2009 Quiz 2 P2 (Time traveling lecturers)] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s2.pdf solution]]
		+

	Genetic Algorithms		Genetic Algorithms
Line 89:		Line 93:
	* [http://courses.csail.mit.edu/6.034f/Examinations/2007q3.pdf 2007 Quiz 3 P1 (Fraud Detection)] [[http://courses.csail.mit.edu/6.034f/Examinations/2007s3.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2007q3.pdf 2007 Quiz 3 P1 (Fraud Detection)] [[http://courses.csail.mit.edu/6.034f/Examinations/2007s3.pdf solution]]
	* [http://courses.csail.mit.edu/6.034f/Examinations/2008q3.pdf 2008 Quiz 3 P1 (Minerals)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s3.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2008q3.pdf 2008 Quiz 3 P1 (Minerals)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s3.pdf solution]]
		+	* [http://courses.csail.mit.edu/6.034f/Examinations/2009q3.pdf 2009 Quiz 3 P1 (Zombies)] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s3.pdf solution]]
		+

	Neural Networks		Neural Networks
Line 107:		Line 113:
	* [http://courses.csail.mit.edu/6.034f/Examinations/2007q3.pdf 2007 Quiz 3 P2 (Forward+Back prop & Tanh)] [[http://courses.csail.mit.edu/6.034f/Examinations/2007s3.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2007q3.pdf 2007 Quiz 3 P2 (Forward+Back prop & Tanh)] [[http://courses.csail.mit.edu/6.034f/Examinations/2007s3.pdf solution]]
	* [http://courses.csail.mit.edu/6.034f/Examinations/2008q3.pdf 2008 Quiz 3 P2 (Adders & Network to Graph matching)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s3.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2008q3.pdf 2008 Quiz 3 P2 (Adders & Network to Graph matching)] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s3.pdf solution]]
		+	* [http://courses.csail.mit.edu/6.034f/Examinations/2008q3.pdf 2009 Quiz 3 P2 (Challenging)] [[http://courses.csail.mit.edu/6.034f/Examinations/2009s3.pdf solution]]
		+

	Self Organizing Maps		Self Organizing Maps
	* [http://courses.csail.mit.edu/6.034f/Examinations/2005f.pdf 2005 Final III.3] [[http://courses.csail.mit.edu/6.034f/Examinations/2005fs.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2005f.pdf 2005 Final III.3] [[http://courses.csail.mit.edu/6.034f/Examinations/2005fs.pdf solution]]
		+

	Support Vector Machines		Support Vector Machines
Line 121:		Line 130:
	* [http://courses.csail.mit.edu/6.034f/Examinations/2007q4.pdf 2007 Quiz 4 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2007s4.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2007q4.pdf 2007 Quiz 4 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2007s4.pdf solution]]
	* [http://courses.csail.mit.edu/6.034f/Examinations/2008q4.pdf 2008 Quiz 4 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s4.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2008q4.pdf 2008 Quiz 4 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s4.pdf solution]]
		+

	Boosting		Boosting
Line 130:		Line 140:
	* [http://courses.csail.mit.edu/6.034f/Examinations/2007q4.pdf 2007 Quiz 4 P2] [[http://courses.csail.mit.edu/6.034f/Examinations/2007s4.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2007q4.pdf 2007 Quiz 4 P2] [[http://courses.csail.mit.edu/6.034f/Examinations/2007s4.pdf solution]]
	* [http://courses.csail.mit.edu/6.034f/Examinations/2008q4.pdf 2008 Quiz 4 P2] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s4.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2008q4.pdf 2008 Quiz 4 P2] [[http://courses.csail.mit.edu/6.034f/Examinations/2008s4.pdf solution]]
		+

	Frames		Frames
	* [http://courses.csail.mit.edu/6.034f/Examinations/2004q2.pdf 2004 Quiz 2 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2004s2.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2004q2.pdf 2004 Quiz 2 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2004s2.pdf solution]]
		+

	Learning and Representation		Learning and Representation
	* [http://courses.csail.mit.edu/6.034f/Examinations/2001fr.pdf 2001 Final-Review P6] [[http://courses.csail.mit.edu/6.034f/Examinations/2001frs.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2001fr.pdf 2001 Final-Review P6] [[http://courses.csail.mit.edu/6.034f/Examinations/2001frs.pdf solution]]
		+

	Concept Learning		Concept Learning
	* [http://courses.csail.mit.edu/6.034f/Examinations/2004f.pdf 2004 Final P5] [[http://courses.csail.mit.edu/6.034f/Examinations/2004fs.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2004f.pdf 2004 Final P5] [[http://courses.csail.mit.edu/6.034f/Examinations/2004fs.pdf solution]]
		+

	Near Miss & Tree Learning		Near Miss & Tree Learning
Line 145:		Line 159:
	* [http://courses.csail.mit.edu/6.034f/Examinations/2006f.pdf 2006 Final V.1] [[http://courses.csail.mit.edu/6.034f/Examinations/2006fs.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2006f.pdf 2006 Final V.1] [[http://courses.csail.mit.edu/6.034f/Examinations/2006fs.pdf solution]]
	* [http://courses.csail.mit.edu/6.034f/Examinations/2007f.pdf 2007 Final V.1] [[http://courses.csail.mit.edu/6.034f/Examinations/2007fs.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2007f.pdf 2007 Final V.1] [[http://courses.csail.mit.edu/6.034f/Examinations/2007fs.pdf solution]]
		+

	Semantic Transition Trees		Semantic Transition Trees
	* [http://courses.csail.mit.edu/6.034f/Examinations/2003q2.pdf 2003 Quiz 2 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2003s2.pdf solution]]		* [http://courses.csail.mit.edu/6.034f/Examinations/2003q2.pdf 2003 Quiz 2 P1] [[http://courses.csail.mit.edu/6.034f/Examinations/2003s2.pdf solution]]
		+

	Misc		Misc