Inspired by the human brain
Like the architecture of the brain, neural networks are layered networks of artificial “neurons”.
Numbers are taken as inputs and output a single number through a simple function.
How a neuron works
A neuron may have multiple inputs.
Its output to the next neuron layer could be a sum of those inputs, or w…..Where i is the number of inputs.
Built for you, trained on your data
Once Cognitiv builds a custom neural network for your brand, it is trained.
Your historical data (such as bid request data) is pushed through the neural network, and the neural network outputs the probability of a click or bid price.
Verification and back-propagation
The neural network observes the actual result (did the impression result in a click?).
Through back-propagation, it adjusts the weights in the network to ensure its output is even more accurate for the next set of inputs.
Enhanced accuracy over time
Just like a human brain learns and hones skill with repetition, the weights in the neural network get more accurate over time.
The ability to improve and identify complex patterns is the most advanced breakthrough in machine learning.