Question 1:
Explain what overfitting is and three different techniques to help avoid it when optimizing deep neural networks.
Question 2:
You have trained five different models on the same set of data and they each get 90% precision. Can you combine these models, without retraining, to get better results? If so explain how. If not, explain why not.
Question 3:
If you are using learners with regularization and AdaBoost underfits the training data, how should you adjust the parameters of AdaBoost or its learners?
Question 4:
An office building with 10 floors has 3 elevators, each of which can hold up to 4 people. Every floor has a pair of call buttons to request up or down service, except the top and bottom floors which have only one button each. When the elevator arrives, a person enters and presses the number of the floor they want. Each elevator can store the floor numbers entered and stops at each floor that is requested. Describe the state and action spaces and calculate their size. Describe a reinforcement learner (reward function and learning method) that can learn to control the elevators, delivering passengers as expected while not wasting energy. Be sure to indicate whether delayed rewards should be used.
Question 5:
Given the three-unit neural network with weights as indicated in which units form products of their weighted inputs rather than sums, write a function for the output value y of node C based on the one input, x, and other network parameters. For example, the input to C would be the product of all incoming weights and associated activations. There are no biases. Unit C is a linear unit, whereas units A and B are sigmoids, φ(z)=(1−e^−z)^−1
If you need solution of above machine learning project then you can contact us at contact@codersarts.com
Comments