Please enable JavaScript.
Coggle requires JavaScript to display documents.
Neural Decoding - Coggle Diagram
Neural Decoding
Likelihood ratio
Cost Models
If the same case of the tiger and bush is represented, the cost for mistaking a tiger for a bush is far greater than vice versa. So there needs to be a cost model for errors in intepretation
-
-
-
Prior Experience
-
If the two things to be distinguished between are a tiger and a bush, it is more likely to see a bush than a tiger, so the response probabilities need to be weighted according to prior experience
Keep taking observations until we are sure that the stimulus crosses a certain threshold value that is to elicit a specific response. Firing rate ramps up until it reaches some threshold upon which it is ready to make a decision
Most efficient statistic and has the most power for a given size of getting the stimulus for a specific response
-
Maximum likelihood
-
-
-
Each neuron's contribution is weighted by its variance. Smaller the variance - the more infomation it has
Population Coding
-
Population Vector
Each neuron adds a component in its preferred direction with a weight that is given by its firing rate
-
Where c_a is the neuron's preferred firing direction and the bracketed term is the normalised firing rate
With enough data, the average will converge and become parallel to the direction of movement.
-
-
-
Tuning Curves
a graph of neuronal response (usually measured in action potentials or spikes per unit time) as a function of a continuous stimulus attribute
Can be used to trace back to the stimulus, for instance by using a cosine tuning curve. However, most of the time the response is too complex to be characterized by a simple tuning curve along which it responds
Poisson Firing
Spikes are produced randomly and independently in each time bin with a probability given by the instantaneous rate
-
-
