what is a statistical model
explicit instantiation of theories
often explains aspects of the theory instead of the whole theory
represents data not phenomena
data
direct observation of phenomena
specific to context it is observed in
predicted by phenomena
phenomena
robust, recurring features of the world
stable generalization of empirical data
e.g. panic disorder
theory
set of linked propositions that explain the phenomenon
abduction from phenomena
deloping a formal model
there is a phenomenon that is to be understood
is arises from a target system
select target system
causal diagram
formal models need to reliably
explain
predict
control
model types
verbal
mathematical —> formulas
diagrams
mechanical models —> DNA
computational models
no ambiguity —>computationally explicit hypotheses!
falsifiable -> can be used to make predictions
often critiqued to be too complex and black box but
complex systems require complex hypotheses
Levels of explanation David Marr
computational theory
what is goal of computation? why appropriate? what is the logic of the strategy=
representation & algorithm
how can theory be implemented
representation for input + output
hardware implementation
how can representation & algorithm be realized physically
Goal: model that represents all three levels
neural networks as models of the brain
input layer
multiple hidden layers
with neurons/units
weights connect to next layer
output layer
example for neural network
input image of digit
output which digit
—> neurons activated in different degrees
—> activation of next neuron depends on incoming weights
deep learning
after calculation of model: feedback what the correct answer was
thats how it learns
convolutional neural etworks
sliding window
moves across image and sums up the results
edge detector: high values mean light and dark are next to each other
CNN filters: random in beginning and learning through feedbacks
DNN as models of the brain
just like gradient in complexity of features across the ventral stream
coarse to fine
deeper levels have later peaks than first levels —> hierarchical just like spatio-temporal cortical dynamics of human visual object recognition
show that recurrence is required —> feedback from deeper layers to primary levels
DNN capture stages of processing in time and space
Failure of DNNs
adversarial examples
no causation
BUT THEY ARE FALSIFIABLE
when inputs are intentionally perturbed in a subtle way to make network make a mistake
e.g. adding noise to pictures —> ostrich
but: maybe ANNs are just more sensitive to adversarial attacks and need another mechanism that makes it more robust
neuroconnectionist research programme
ANNs as computational language for expressing falsifiable theories about brain computation
Reductionist POV:
reductionist pov
theory -> hypothesis -> prediction -> test on data -> theory falisified -> theory rejected
but in reality there is often a set of hypotheses and not the whole theory needs to be rejected!
BELT is tested
neuroconnectionism
uses biologically inspired ANNs to model human behavioral and neural data
solution: model by neuroconnectionist research program
core
progressive degenerative:
empirical data
Belt:
model by neuroconnectionist research program core
core: fixed background assumptions that are not challenged from within the programme and not substantially altered without abandoning the programme
model by neuroconnectionist research program progressive degenerativ
progressive programmes: generate successful predictions
degenerative: do not generate novel predictions and fail to account for empirical data
model by neuroconnectionist research program empirical data
empirical data: used to test belt hypotheses
model by neuroconnectionist research program belt
Belt: auxiliary hypotheses that are experimentally tested and subject to change when new evidence comes up
testing the belt
behavioral agreement
agreement with neural data -> representational similarity analysis
in silico electrophysiology —> lesion studies
developmental agreement —> learning vs. training development
Last changed5 months ago