1
SUPPLEMENTARY FIGURES
2
1
3
Fig. S1
4
Including an additional parameter (spatial frequency) along with orientation in a single-
5
feature map development process driven by a Spatial learning rule.
6
7
(A) Four examples are shown of the single feature, 2-parameter stimulus. (B) Learned receptive
8
fields of some sample cells show the micro-organization of the parameters (spatial frequency and
9
orientation). RFs are shown in retinal space. (C) Mature orientation (OR) map after 210K
10
stimulation steps. Arrows show gradient direction and magnitude. White dots mark locations
11
where OR gradient is roughly orthogonal to gradient in the spatial frequency (SF) map shown in
12
(C). (D) Spatial frequency map corresponding to the orientation map in (B). Here arrows indicate
13
SF map gradients. White dots are same as in (C). (E) Gradient arrows from (C) and (D) are
14
shown together wherever both lie within the 20th to 80th percentile range of gradient magnitudes
15
(gradients outside that range were sometimes unreliable). Green dots mark sites where gradient
16
directions in the OR and SF maps are close to orthogonal (90 ± 30°). (F) Histogram of the angles
17
between OR and SF gradients from (E) shows peaks near +90° and -90°, indicating the two
18
gradients tend to be orthogonal.
19
2
20
21
22
23
24
Fig. S2
25
Developing a map under Hybrid learning when stimulated by a more complex multi-bar
26
stimulus.
27
28
(A) Four examples are shown of the more complex stimulus. Different random
29
locations/orientations were generated for each stimulus. (B) Learned receptive fields of some
30
sample cells show the RF is dominated by a single oriented feature, surrounded by low contrast
31
noise resulting from the extraneous stimuli. (C, D) Orientation map and combined orientation/RF
32
center (“Contortion”) plot are comparable to the single-feature Hybrid outcomes using a simple
33
stimulus (see Fig. 4).
3