Appeal No. 97-0507
Application 08/202,609
operation associated with known data before the neural network
essentially is “turned loose” upon raw, new data. Changing the
weighting functions associated with a training operation of a
neural operation does not change the structure or the hardware of
the associated electronics of the actual device. The electrical
values do change but the actual electrical circuits processing
the trainable data values do not change. The mathematical values
or weights change during training in any neural network that is
“trainable.” The act of changing such weighting functions occurs
inherently in any neural network that is “trainable.” The
specific weighting functions changed in accordance with the last
clause of the body of claim 25 on appeal are not recited in the
claim to be changed from one form to another. It is inherent
within the nature of the devices of the prior art relied upon by
the examiner that, for example, the seismic data traces processed
by McCormack may have yielded identical weighting functions to
those only generically recited in this clause of claim 25 on
appeal. The “trainability” recited in the last clause of the
body of claim 25 on appeal affecting the sum function and
transfer function recited in two instances in the first clause of
the body of claim 25 on appeal are inherent properties of the
11
Page: Previous 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Next
Last modified: November 3, 2007