It is also interesting to examine the behaviour of a given model as the size of the data set is varied, as shown in Figure 1.6. We see that, for a given model complexity, the over-fitting problem become less severe as the size of the data set increases. Another way to say this is that the larger the data set, the more complex (in other words more flexible) the model that we can afford to fit to the data.
Also, there is something rather unsatisfying about having to limit the number of parameters in a model according to the size of the available training set. It would seem more reasonable to choose the complexity of the model according to the complexity of the problem being solved. We shall see that the least squares approach to finding the model parameters represents a specific case of maximum likelihood (discussed in Section 1.2.5), and that the over-fitting problem can be understood as a general property of maximum likelihood. By adopting a Bayesian approach, the over-fitting problem can be avoided. We shall see that there is no difficulty from a Bayesian perspective in employing models for which the number of parameters greatly exceeds the number of data points. Indeed, in a Bayesian model the effective number of parameters adapts automatically to the size of the data set.
One technique that is often used to control the over-fitting phenomenon in such cases is that of regularization, which involves adding a penalty term to the error function (1.2) in order to discourage the coefficients from reaching large values.
Here we simply note that, if we were trying to solve a practical application using this approach of minimizing an error function, we would have to find a way to determine a suitable value for the model complexity. The results above suggest a simple way of achieving this, namely by taking the available data and partitioning it into a training set, used to determine the coefficients w, and a separate validation set, also called a hold-out set, used to optimize the model complexity (either M or λ). In many cases, however, this will prove to be too wasteful of valuable training data, and we have to seek more sophisticated approaches.
The quantity p(D|w) on the right-hand side of Bayes’ theorem is evaluated for the observed data set D and can be viewed as a function of the parameter vector w, in which case it is called the likelihood function. It expresses how probable the observed data set is for different settings of the parameter vector w. Note that the likelihood is not a probability distribution over w, and its integral with respect to w does not (necessarily) equal one.
In Gaussian Distribution, on average the maximum likelihood estimate will obtain the correct mean but will underestimate the true variance by a factor (N − 1)/N.
Information Theory:
The form of h(·) can be found by noting that if we have two events x and y that are unrelated, then the information gain from observing both of them should be the sum of the information gained from each of them separately, so that h(x, y) = h(x) + h(y). Two unrelated events will be statistically independent and so p(x, y) = p(x)p(y). From these two relationships, it is easily shown that h(x) must be given by the logarithm of p(x).
This relation between entropy and shortest coding length is a general one. The noiseless coding theorem (Shannon, 1948) states that the entropy is a lower bound on the number of bits needed to transmit the state of a random variable.
We have introduced the concept of entropy in terms of the average amount of information needed to specify the state of a random variable. In fact, the concept of entropy has much earlier origins in physics where it was introduced in the context of equilibrium thermodynamics and later given a deeper interpretation as a measure of disorder through developments in statistical mechanics.
