Comment for Midterm Examination

Regrading Policy for the midterm exam

  1. Only for obvious mistake of the grader such as miscounting lost points.
  2. Only when you have explicitly stated the answers correctly but still lost point >= 4. (Please DO NOT argue "implied answers, etc.".)
    1. Please don't come for point loss less than 4
  3. Regrading will be done on the entire exam, which may reduce your grade. No exception.
  4. No argument against the grading standard given below.
  5. Regrading request will be accepted until Sunday, October 27, 2002.  No exception.

Grading Standard for the midterm exam

1.g.  Feedforward model of Hubel & Wiesel (lecture note 05 p34)
V1 cells receive inputs from LGN cells arranged along a given orientation. (1)
Orientation tuned cell.(2)
Superimposed alignment of center-surround cells along the selected orientation. (4)
Superimposed input to receptive field (2)
Correct Diagram (3)

1.h.  Retinotopic map
topographic preservation of retinal input and v1 input (3)
radio-active tracer used to see the stimulus patterns of a moneky's brain after shown stimulus on screen (2)

source separation of post-processing because each electrode may record many neurons around it.(4)
one electrode may records signals of many neurons around it.(4)
isolation of sources.(4)
no knowledge of exact location of electrode.(2)
invasive technology.(1)

distinction of order of stimulus appearance.(3)
bias of answer yes/no persistently in case of uncertainty.(3)

(4) {
PET: emission of two photons travelingin opposite direction gives better clue to generate less scattered brain imaging.
}Correct answer
(2) {PET: two photons; SPECT: one photos}
(2) {PET: better accuracy.}
(1) {PET: cyclotron; SPECT: tracers easily created.}
(1) {SPECT: wide range of application}
(1) {SPECT: tracers decaly slowly, so experiment is slow.}
(1) {any set of two differences}

(3) {detection of change or gradient of magnetic moment due to deoxygenation}
(1) { oxygen level }

(3) {increase in neural activity -> more deoxygenation -> increase in magnetic moment}
(2) {increase in neuroal activity -> more fresh blood (more oxygen) -> more deoxygenation }

to open a hole in a skull and to shine a light to videotape reflectance properties of neurons due to change of activities (5)

McCulloch Pitts neuron(5)

w = 1, k-1 < theta <= k (2)
example acceptable

w=1, 0 < theta <= 1 (2)

n (1)
max( m(i) ) <= j <= summation ( m(i) ) : as i go through 1 to k (1)

1 (1)

k (4)

n input one-to-many linked to j iput neurons or n input neurons (n)
j input neurons many-to-one linked to k hidden layer neurons (5)
k hidden layer neurons many-to-one linked to 1 decision layer neuron (5)

3.c.1 no
3.c.2 no
3.c.3 b, c
3.c.4 L+R+C (=S)
3.c.5 S < T
3.c.6 S >= T
3.c.7 L(b) > L(a)
for c: S >=T; for d; S < T
L(c) > L(d)
L(c) = L(a); L(d) = L(b)
==>  L(a) > L(b) <== contradiction with 3.c.7