Fwd: [ECS 170] Assignment 3 Question

[dhye]

---------- Forwarded message ----------
From: Alex Tsui <alext...@gmail.com>
Date: Thu, May 31, 2012 at 11:22 AM
Subject: Re: [ECS 170] Assignment 3 Question
To: Daihua ye <dh...@ucdavis.edu>

On Thu, May 31, 2012 at 10:38 AM, Ian Davidson <davi...@cs.ucdavis.edu> wrote:

---------- Forwarded message ----------
From: "dhye" <dh...@ucdavis.edu>
Date: May 31, 2012 10:33 AM
Subject: [ECS 170] Assignment 3 Question
To: "Ian Davidson" <ian.n.d...@gmail.com>, "Ian Davidson" <davi...@cs.ucdavis.edu>, "Alex Tsui" <alext...@gmail.com>

Hi Professor,

For 1, we need implement -train and -test option to test data. When we have -train option, do we need provide the path of files from Female and Male?

For example:

java teamname -tain Female/*.txt Male/*.txt

or just

java teamname -tain

problem in the second approach: the program dont know what is the location of test files if we don't provide the path to files

It's okay if you provide the path to the files, right? Just provide the path then - it shouldn't be a problem.
 

For 2, "cross-validation" is mean randomly use one fold as train and another fold as test? The train and test fold must be different?

Yes, the train and test folds must be different. The training set should have 4 folds and the test fold is the remaining 1.
 

For 3, the format of report: machine readable format? Is it just a table of prediction and confidence of my neural network?

Also, what is "confidence neural network"? do you have any example for explain that? Is every test case need the confidence neural network?

For example:

Test:

{MALE}, 0.8

{FEMALE}, 0.9

...

Yes, just give me something that looks like that example and it should be okay.
 

For 4, The results, do we need to write code for implement the images in the results? or just writeup in the result?

Professor explained that you can take the weights and scale them to integer values in 0-255, then save to file. Then you can use View.java that I uploaded to smartsite to visualize those files.
 

Thanks.

--

Best,

Daihua Ye

College of Letter and Science,

University of California, Davis

B.S. Computer Science

dh...@ucdavis.edu | (408)-386-7053

My Portfolio

--

Best,

Daihua Ye

College of Letter and Science,

University of California, Davis

B.S. Computer Science

dh...@ucdavis.edu | (408)-386-7053

My Portfolio