Machine Learning: Question Set – 17

Machine Learning: Question Set – 17

When does regularization enter the picture in Machine Learning?

Regularization is required when the model begins to underfit or overfit. Regularization is a regression that directs or regularizes the estimations of the coefficients towards zero.


To minimize overfitting, it decreases flexibility and discourages learning in a model. The model’s complexity decreases, and it gets more predictive.

Over fitted and regularization curve
Over fitted and Regularized curve

You are provided a data collection with missing values ranging from 1 standard deviation from the mean. How much of the data would be unaltered?

It is assumed that the data is distributed across a mean, i.e. the data is distributed across an average. As a result, we can assume that it is a normal distribution. In a normal distribution, approximately 68 percent of data is within one standard deviation of averages such as mean, mode, or median. This suggests that around 32% of the data is unaffected by missing values.

Is a high variance in data beneficial or bad?

Higher variance indicates that the data spread is wide and the feature contains a diverse set of data. High variance in a feature is usually regarded as poor quality.

What is the trade-off relationship between bias and variance?

Bias and variance are both errors. Bias is a mistake caused by incorrect or too simplistic assumptions in the learning algorithm. It may cause the model to underfit the data, making it difficult to achieve high predicted accuracy and generalize knowledge from the training set to the test set.

Variance is an error caused by the learning algorithm’s overcomplexity. It causes the algorithm to be extremely sensitive to high levels of variation in the training data, which can cause the model to overfit the data.

To limit the amount of errors as much as possible, we must trade off bias and variance.

In Machine Learning, what is a model selection?

Model selection refers to the process of selecting models from among various mathematical models that are used to define the same data. Model learning is applied to the fields of statistics, data mining, and machine learning.

What is the distinction between generative and discriminative models?

A generative model will learn data categories, whereas a discriminative model would just learn the distinction between different data categories.

On classification challenges, discriminative models will generally outperform generative models.

What is the distinction between Type I and Type II errors?

The first type of error is a false positive, while the second type of error is a false negative. In a nutshell, Type I mistake implies stating something happened when it didn’t, but Type II error means claiming nothing is occurring when something is.

One creative approach to think about this is to imagine Type I error as telling a man he is pregnant, whereas Type II error as telling a pregnant woman she is not carrying a baby.

What is the difference between bias and variance in a machine learning model?

Bias: Bias occurs in a machine learning model when the predicted values differ from the actual values. A model with low bias has forecast values that are very close to the actual ones. Underfitting: When an algorithm is biased, it may miss significant relationships between features and intended outputs.

Variance: The amount by which the target model changes when trained with different training data is referred to as variance. The variance of a good model should be kept to a minimum. Overfitting: When an algorithm has a high variance, it may model the random noise in the training data rather than the intended results.

What exactly do you mean by the Reinforcement Learning technique?

Reinforcement learning is a machine learning algorithm technique. It entails an agent interacting with its surroundings by performing activities and discovering faults or rewards. Different software and computers use reinforcement learning to find the best suitable behavior or path to take in a given situation. It normally learns by rewarding or punishing itself for every action it does.

Reinforcement Learning
Reinforcement Learning

What are the various Algorithm Methods in Machine Learning?

The various sorts of algorithm strategies used in machine earning are as follows:

  • Supervised Learning
  • Semi-supervised Learning
  • Unsupervised Learning
  • Reinforcement Learning
  • Transduction

How would you deal with a dataset that has a high variance?

We could utilize the bagging technique to handle high variance datasets. The tagging method divides the data into subgroups using sampling from random data. Following the splitting of the data, random data is utilized to construct rules using a training method.

The polling technique is then used to combine all of the model’s projected results.

What does the term “overfitting” in machine learning mean?

In machine learning, overfitting occurs when a statistical model describes random error or noise rather than the underlying relationship. Overfitting is commonly found when a model is overly complex.

It occurs as a result of having too many parameters pertaining to the amount of training data types. The model’s performance is poor since it has been overfitted.

Regularization helps to overcome the issue of overfitting by controlling the magnitude of parameters.

Two popular techniques are:

  • Regularization using Lasso regression
  • Regularization using Ridge regression

Additional Reading: Prevent overfitting with

Leave a Reply

Your email address will not be published. Required fields are marked *