Looking for insight into solving below HW questions.

IlI. lConsider the multiperiod casino betting competition of Session 8, where a player has a 52% probability of winning each round of play, there are 50 total rounds of play, and in each round of play a constant fraction of the total wealth is wagered. Which of the following is true? (Select all correct answers.) {a} The larger the bet fraction, the higher the expected wealth at the end of the competition. [b] The larger the bet fraction, the higher the median wealth at the end of the com petition. (c) The larger the bet fraction, the greater the risk for the player. IV. Neural networks are a predictive analytical tool that have been used to generate high quality predictions in a variety of diverse application areas. The general idea is that the independent variables. are combined with each other to produce intermediate variables (in several layers], which are then combined with each other to produce a prediction. 4 The parameters of the functions that dene the intermediate variables in terms of the inputs are learnedftrained in order to obtain the best t. [0hr example: The coefcients ofa linear model are its parameters) Consider the problem of predicting the genre of a music clip that is 10 seconds in length. With a sampling rate of 20 kHz {i_e_, 20,01?!) samples. per second}, this means that the number of independent variables is 203,000. There are two neural network based predictive approaches available: FMN: A feedforward neural network [FNN] for generating predictions is specied by 4.2 million model parameters that must be learned from data. CDAE: A convolutional denoising autoencoder (CDAE) neural network, on the other hand, is specied by 37 thousand model parameters. Which of the following is true? (Select all correct answers.) {a} If a limited amount of training data is available, the CDAE approach may provide better in-saInple predictive performance on the training data because it has fewer parameters. [b] If a limited amount of training data is available, the CDAE approach may provide better outof-sample performance on test data because it has fewer parameters. (c) If a my large amount of training data is available, this may make the FNN approach more attractive, because it may be able to detect more intricate patterns than the CDAE' approach. [d] The FNN approach is leg vulnerable to overtting than the CDAE approach because it involves more model parameters