Glossary of common Machine Learning, Statistics and Data Science terms

Word

Description

So, the accuracy is the ratio of correctly predicted classes to the total classes predicted. Here, the accuracy will be:

Features:

1. It is computationally efficient and has little memory requirements
2. It is invariant to diagonal rescaling of the gradients
3. Adam works well in practice as compared to other stochastic optimization methods

1. Speed − Spark helps to run an application in Hadoop cluster, up to 100 times faster in memory, and 10 times faster when running on disk
2. Spark supports popular data science programming languages such as R, Python, and Scala
3. Spark also has a library called MLlIB which includes basic machine learning including classification, regression, and clustering

For example, we can predict the value for the next time step (t+1) given the observations at the last two time steps (t-1 and t-2). As a regression model, this would look as follows:

X(t+1) = b0 + b1*X(t-1) + b2*X(t-2)

Since the regression model uses data from the same input variable at previous time steps, it is referred to as an autoregression.

Word

Description

• Bagging meta-estimator
• Random Forest

To gain a better understanding about bar charts, refer here.

Bayes’ theorem is used to calculate the conditional probability. Conditional probability is the probability of an event ‘B’ occurring given the related event ‘A’ has already occurred.

For example, Let’s say a clinic wants to cure cancer of the patients visiting the clinic.

A represents an event “Person has cancer”

B represents an event “Person is a smoker”

The clinic wishes to calculate the proportion of smokers from the ones diagnosed with cancer.

To do so use the Bayes’ Theorem (also known as Bayes’ rule) which is as follows:
To understand Bayes’ Theorem in detail, refer here.

Following are these component :

1. Bias error is useful to quantify how much on an average are the predicted values different from the actual value
2. Variance on the other side quantifies how are the prediction made on same observation different from each other

A high bias error means we have a under-performing model which keeps on missing important trends. A high variance model will over-fit on your training population and perform badly on any observation beyond training. In order to have a perfect fit in the model, the bias and variance should be balanced which is bias variance trade off.

Binomial Distribution is applied only on discrete random variables. It is a method of calculating probabilities for experiments having fixed number of trials.

Binomial distribution has following properties:

1. The experiment should have finite number of trials
2. There should be two outcomes in a trial: success and failure
3. Trials are independent
4. Probability of success (p) remains constant

For a distribution to qualifying as binomial, all of the properties must be satisfied.

So, which kind of distributions would be considered binomial? Let’s answer it using few examples:

1. Suppose, you need to find the probability of scoring bull’s eye on a dart. Can it be called as binomial distribution? No, because the number of trials isn’t fixed. I could hit the bull’s eye on the 1st attempt or 3rd attempt or I might not be able to hit it at all. Therefore, trials aren’t fixed.
2. A football match can have resulted in 3 ways: Win, Lose or Draw. Thus, if we are asked to find the probability of winning in this case, binomial distribution cannot be used because there are more than two outcomes.
3. Tossing a fair coin 20 times is a case of binomial distribution as here we have finite number of trials 20 with only two outcomes “Head” or “Tail”. These trials are independent and probability of success is 1/2 across all trials.

The formula to calculate probability using Binomial Distribution is:

P ( X = r ) = nCr (pˆr)* (1-p) * (n-r)

where:
n : No. of trials
r : No. of success
p : the probability of success
1 – p : Probability of failure
nCr : binomial coefficient given by n!/k!(n-k)!

• AdaBoost
• GBM
• XGBM
• LightGBM
• CatBoost

To learn more about boosting algorithms, refer here.

Some of the inferences that can be made from a box plot:

• Median: Middle quartile marks the median.
• Middle box represents the 50% of the data
• First quartile: 25% of data falls below these line
• Third quartile: 75% of data falls below these line.

Word

Description

For example: Classification Algorithms like Logistic Regression, Decision Tree, K-NN, SVM etc.

Clustering is an unsupervised learning method used to discover the inherent groupings in the data.  For example: Grouping customers on the basis of their purchasing behaviour which is further used to segment the customers. And then the companies can use the appropriate marketing tactics to generate more profits.

Example of clustering algorithms: K-Means, hierarchical clustering, etc.

• Pedestrians, cars, road detection in smart (self-driving) cars
• Object recognition
• Object tracking
• Motion analysis
• Image restoration

Let’s say you had two movie reviewers rank a set of 5 movies:

The ranks given by the reviewer 1 are ordered in ascending order, this way we can compare the rankings given by both the reviewers.

Concordant Pair – 2 entities would form a concordant pair if one of them is ranked higher than the other consistently. For example, in the table above B and D form a concordant pair because B has been ranked higher than D by both the reviewers.

Discordant Pair – C and D are discordant because they have been ranked in opposite order by the reviewers.

Concordant Pair or Discordant Pair ratio = (No. of concordant or discordant pairs) / (Total pairs tested)

Convex functions play an important role in many areas of mathematics. They are especially important in the study of optimization problems where they are distinguished by a number of convenient properties.

The most widely used correlation coefficient is Pearson Coefficient. Here is the mathematical formula to derive Pearson Coefficient.

Here,

• h(x) is the prediction
• y is the actual value
• m is the number of rows in the training set

Let us understand it with an example:

So let’s say, you increase the size of a particular shop, where you predicted that the sales would be higher. But despite increasing the size, the sales in that shop did not increase that much. So the cost applied in increasing the size of the shop, gave you negative results. So, we need to minimize these costs. Therefore we make use of cost function to minimize the loss.

Where,

x = the independent variable

y = the dependent variable

n = number of data points in the sample

x bar = the mean of the independent variable x

y bar = the mean of the dependent variable y

A positive covariance means the variables are positively related, while a negative covariance means the variables are inversely related.

• Leave one out cross validation (LOOCV)
• k-fold cross validation
• Stratified k-fold cross validation
• Adversarial validation

Word

Description

1. Mining for frequent patterns
2. Mining for associations
3. Mining for correlations
4. Mining for clusters
5. Mining for predictive analysis

Data Mining is done for purposes like Market Analysis, determining customer purchase pattern, financial planning, fraud detection, etc

For instance, replacing a variable x by the square root of x

• Dict of 1D ndarrays, lists, dicts, or Series
• 2-D numpy.ndarray
• Structured or record ndarray
• A series
• Another DataFrame

The algorithm has two important features:

• distance
• the minimum number of points required to form a dense region

The steps involved in this algorithm are:

• Beginning with an arbitrary starting point it extracts the neighborhood of this point using the distance
• If there are sufficient neighboring points around this point then a cluster is formed
• This point is then marked as visited
• A new unvisited point is retrieved and processed, leading to the discovery of a further cluster or noise
• This process continues until all points are marked as visited

The below image is an example of DBScan on a set of normalized data points:

Here the lines separating each class are decision boundaries.

Decision tree is a type of supervised learning algorithm (having a pre-defined target variable) that is mostly used in classification problems. It works for both categorical and continuous input & output variables. In this technique, we split the population (or sample) into two or more homogeneous sets (or sub-populations) based on most significant splitter / differentiator in input variables.

Read more here.

For example, the diagram below shows the health score of a patient from range 0 to 60. Nine deciles split the patients into 10 groups

For example, in a sample of size 10 with mean 10, 9 values can be arbitrary but the 10th value is forced by the sample mean. So, we can choose any number for 9 values but the 10th value must be such that the mean is 10. So, the degree of freedom in this case will be 9.

• It helps in data compressing and reducing the storage space required
• It fastens the time required for performing same computations
• It takes care of multicollinearity that improves the model performance. It removes redundant features
• Reducing the dimensions of data to 2D or 3D may allow us to plot and visualize it precisely
• It is helpful in noise removal also and as result of that we can improve the performance of models

It can be easily installed using the following code from the R console:

install.packages("dplyr")

Word

Description

For example, in XGBoost, as you train more and more trees, you will overfit your training dataset. Early stopping enables you to specify a validation dataset and the number of iterations after which the algorithm should stop if the score on your validation dataset didn’t increase.

The steps involved in EDA are:

1. Variable IdentificationIn this step, we identify the data type and category of variables
   
2. Univariate analysis
3. Multivariate analysis

Refer here for a comprehensive guide to doing EDA.

• It extracts data from the source systems
• It enforces data quality and consistency standards
• Delivers data in a presentation-ready format

This data can be used by application developers to build applications and end users for making decisions.

1. AUC
2. ROC score
3. F-Score
4. Log-Loss

Word

Description

• There is linear relationship
• There is no multicollinearity
• It includes relevant variables into analysis
• There is true correlation between variables and factors

There are different types of methods used to extract the factor from the data set:

1. Principal Component Analysis
2. Common factor analysis
3. Image factoring
4. Maximum likelihood method

Suppose we have three documents:

• John likes to watch movies.
• Mary likes movies too.
• John also likes football.

Now we can convert this to vector using hashing.

The array form for the same will be:

PCA is one of the most popular feature reduction techniques, where we combine correlated variables to reduce the features.

This can be done by either filtering out less useful features or by combining features to make a new one.

Refer here.

Here are the major features of Apache Flume:

• Flume is a flexible tool as it allows to scale in environments with as low as five machines to as high as several thousands of machines
• Apache Flume provides high throughput and low latency
• Apache Flume has a declarative configuration but provides ease of extensibility
• Flume in Hadoop is fault tolerant, linearly scalable and stream oriented

Frequentist Statistics tests whether an event (hypothesis) occurs or not. It calculates the probability of an event in the long run of the experiment (i.e the experiment is repeated under the same conditions to obtain the outcome).

Here, the sampling distributions of fixed size are taken. Then, the experiment is theoretically repeated infinite number of times but practically done with a stopping intention. For example, I perform an experiment with a stopping intention in mind that I will stop the experiment when it is repeated 1000 times or I see minimum 300 heads in a coin toss. Read more here.

F measure = 2 x (Recall × Precision) / ( β² × Recall + Precision )

Word

Description

Instead of the input, forget, and output gates in the LSTM cell, the GRU cell has only two gates, an update gate z, and a reset gate r. The update gate defines how much previous memory to keep, and the reset gate defines how to combine the new input with the previous memory.

It can be easily installed using the following code from the R console:

install.packages("ggplot2")

The main features of Go are:

• Memory safety
• Garbage collection
• Structural typing

The compiler and other tools originally developed by Google are all free and open source. To read further on the Go language, refer here.

With regard to a machine learning algorithm, a good fit is when the error for the model on the training data as well as the test data is minimum. Over time, as the algorithm learns, the error for the model on the training data goes down and so does the error on the test dataset. If we train for too long, the performance on the training dataset may continue to decrease because the model is overfitting and learning the irrelevant detail and noise in the training dataset. At the same time the error for the test set starts to rise again as the model’s ability to generalize decreases.

So the point just before the error on the test dataset starts to increase where the model has good skill on both the training dataset and the unseen test dataset is known as the good fit of the model.

• On the basis of data ingestion:
  1. Full Batch Gradient Descent Algorithm
  2. Stochastic Gradient Descent Algorithm

In full batch gradient descent algorithms, we use whole data at once to compute the gradient, whereas in stochastic we take a sample while computing the gradient.

• On the basis of differentiation techniques:
  1. First order Differentiation
  2. Second order Differentiation

Word

Description

• Hadoop clusters work and keeps multiple copies to ensure reliability of data. A maximum of 4500 machines can be connected together using Hadoop
• The whole process is broken down into pieces and executed in parallel, hence saving time. A maximum of 25 Petabyte (1 PB = 1000 TB) data can be processed using Hadoop
• In case of a long query, Hadoop builds back up data-sets at every level. It also executes query on duplicate datasets to avoid process loss in case of individual failure. These steps makes Hadoop processing more precise and accurate
• Queries in Hadoop are as simple as coding in any language. You just need to change the way of thinking around building a query to enable parallel processing

HMM are widely used for pattern recognition in speech recognition, part-of-speech tagging, handwriting recognition, and reinforcement learning.

Hierarchical clustering, as the name suggests is an algorithm that builds hierarchy of clusters. This algorithm starts with all the data points assigned to a cluster of their own. Then two nearest clusters are merged into the same cluster. In the end, this algorithm terminates when there is only a single cluster left.

The results of hierarchical clustering can be shown using dendrogram. The dendrogram can be interpreted as:

Read more here.

Histograms are widely used to determine the skewness of the data. Looking at the tail of the plot, you can find whether the data distribution is left skewed, normal or right skewed.

• Indexing to provide acceleration
• Different storage types such as plain text, RDFile, HBase, ORC, and others
• Metadata storage in a relational database management system, significantly reducing the time to perform semantic checks during query execution
• Operating on compressed data stored into the Hadoop ecosystem

For detailed information, refer here.

For instance, if I divide my data in two parts – 7:3 and use the 70% to train the model, and other 30% to check the performance of my model, the 30% data is called the holdout sample.

Visit here for more details.

Some keys points about the hyperparameters are:

• They are often used in processes to help estimate model parameters.
• They are often manually set.
• They are often tuned to tweak a model’s performance

Number of trees in a Random Forest, eta in XGBoost, and k in k-nearest neighbours are some examples of hyperparameters.

Most of the time it’s basically a normal plane, but in some special cases, like in Support Vector Machines, where classifications are performed with an n-dimensional hyperplane, the n can be quite large.

Word

Description

For example,

If the data is as below

The second row contains a missing value, so to impute it we use mean of all ages, i.e.

Word

Description

• Multiple dispatch: providing the ability to define function behavior across many combinations of argument types.
• Good performance, approaching that of statically-compiled languages like C
• Built-in package manager
• Designed for parallelism and distributed computation
• Free and open source

Word

Description

It is a type of unsupervised algorithm which solves the clustering problem. It is a procedure which follows a simple and easy way to classify a given data set through a certain number of clusters (assume k clusters). Data points inside a cluster are homogeneous and heterogeneous to peer groups.

Following are some important features of Keras:

• User friendliness
• Modularity
• Easy extensibility
• Work with Python

K nearest neighbors is a simple algorithm that stores all available cases and classifies new cases by a majority vote of its k neighbors. The case being assigned to the class is most common amongst its K nearest neighbors measured by a distance function.

These distance functions can be Euclidean, Manhattan, Minkowski and Hamming distance. First three functions are used for continuous function and fourth one (Hamming) for categorical variables. If K = 1, then the case is simply assigned to the class of its nearest neighbor. At times, choosing the value for K can be a challenge while performing KNN modeling.

• Read more here.

Kurtosis is defined as the thickness (or heaviness) of the tails of a given distribution. Depending on the value of kurtosis, it can be classified into the below 3 categories:

• Mesokurtic: The distribution with kurtosis value equal to 3. A random variable which follows a normal distribution has a kurtosis value of 3
• Platykurtic: If the kurtosis is less than 3. In this, the given distribution has thinner tails and a lower peak than a normal distribution
• Leptykurtic: When the kurtosis value is greater than 3. In this, the given distribution has fatter tails and a higher peak than a normal distribution

Word

Description

Labeled data are usually more expensive to obtain than the raw unlabeled data because preparation of the labelled data involves manual labelling every piece of unlabeled data.

Labeled data is required for supervised learning algorithms.

Lasso regression performs L1 regularization, i.e. it adds a factor of sum of absolute value of coefficients in the optimization objective. Thus, lasso regression optimizes the following:

Objective = RSS + α * (sum of absolute value of coefficients)

Here, α (alpha) works similar to that of ridge and provides a trade-off between balancing RSS and magnitude of coefficients. Like that of ridge, α can take various values. Let’s iterate it briefly here:

1. α = 0 : Same coefficients as simple linear regression
2. α = ∞ : All coefficients zero (same logic as before)
3. 0 < α < ∞ : coefficients between 0 and that of simple linear regression

In the plot below, for each time instance, the speed trend is shown and the points are connected to display the trend over time.

This plot is for a single case. Line charts can also be used to compare changes over the same period of time for multiple cases, like plotting the speed of a cycle, car, train over time in the same plot.

The best way to understand linear regression is to relive this experience of childhood. Let us say, you ask a child in fifth grade to arrange people in his class by increasing order of weight, without asking them their weight! What do you think the child will do? He / she would likely look (visually analyze) at the height and build of people and arrange them using a combination of these visible parameters. This is linear regression in real life. The child has actually figured out that height and build would be correlated to the weight by a relationship, which looks like the equation below.

Y=aX+b

where:

• Y – Dependent Variable
• a – Slope
• X – Independent variable
• b – Intercept

These coefficients a and b are derived based on minimizing the sum of squared difference of distance between data points and regression line.

Look at the below example. Here we have identified the best fit line having linear equation y=0.2811x+13.9. Now using this equation, we can find the weight, knowing the height of a person.

Mathematically, log loss for two classes is defined as:

where, y is the class label and p is the predicted probability.

• Time series predictions
• Speech recognition
• Rhythm learning
• Handwriting recognition

To learn further on LSTM, refer here.

Word

Description

Features of Mahout:

• Mahout offers a framework for doing data mining tasks on large volumes of data

• Mahout lets applications to analyze large sets of data effectively and in quick time

• It also offers distributed fitness function capabilities for evolutionary programming

• It includes several MapReduce enabled clustering implementations such as k-means, fuzzy k-means, Dirichlet, and Mean-Shift

A MapReduce framework is usually composed of three operations:

1. Map: each worker node applies the map function to the local data, and writes the output to a temporary storage. A master node ensures that only one copy of redundant input data is processed.
2. Shuffle: worker nodes redistribute data based on the output keys (produced by the map function), such that all data belonging to one key is located on the same worker node.
3. Reduce: worker nodes now process each group of output data, per key, in parallel.

To learn more about MapReduce, visit here.

When two or more products are purchased, Market Basket Analysis is done to check whether the purchase of one product increases the likelihood of the purchase of other products. This knowledge is a tool for the marketers to bundle the products or strategize a product cross sell to a customer.

Suppose the total sale is 100$, this total can be broken into sub components i.e. 60$ base sale, 20$ pricing, 18$ may be distribution and 2$ might be due to promotional activity. These numbers can be achieved using various logical methods. Every method can give a different break up. Hence, it becomes very important to standardize the process of breaking up the total sales into these components. This formal technique is formally known as MMM or Market Mix Modeling.

For instance, if you have the marks of students from a class, and you asked about how good is the class performing. It would be irrelevant to say the marks of every single student, instead, you can find the mean of the class, which will be a representative for class performance.
To find the mean, sum all the numbers and then divide by the number of items in the set.

For example, if the numbers are 1,2,3,4,5,6,7,8,8 then the mean would be 44/9 = 4.89.

To calculate the median for a set of numbers, follow the below steps:

1. Arrange the numbers in ascending or descending order
2. Find the middle value, which will be n/2 (where n is the numbers in the set)

Objectives of MIS:

• To improve decision-making, by providing up-to-date, accurate data on a variety of organizational assets
• To correlate multiple data points in order to strategize ways to improve operations

• Microsoft Azure Machine Learning Studio
• AWS Machine Learning
• IBM Watson Machine Learning
• Google Cloud Machine Learning Engine
• BigML

Mode can be calculated using following steps:

• Count the number of time each value appears
• Take the value which appears the most

Let us understand it with an example:

Suppose we have a dataset having 10 data points, listed below:

4,5,2,8,4,7,6,4,6,3

So now we will calculate the number of times each value has appeared.

So we see that the value 4 is repeating the most, i.e., 3 times. So, the mode of this dataset will be 4.

• Exploratory Data Analysis
• Scientific Methods

Some of the criteria for selecting the model can be:

• Akaike Information Criterion (AIC)
• Adjusted R2
• Bayesian Information Criterion (BIC)
• Likelihood ratio test

For example, if the target is to predict the quality of a product, which can be Excellent, good, average, fair, bad. In this case, the variable has 5 classes, hence it is a 5-class classification problem.

For example, we can perform bivariate analysis of combination of two continuous features and find a relationship between them.

Consider the example – for a given set of details about a student’s interests, previous subject-wise score etc, you want to predict the GPA for all the semesters (GPA1, GPA2, …. ). This problem statement can be addressed using multivariate regression since we have more than one dependent variable.

Word

Description

ChatBot is a classic example of NLP, where sentences are first processed, cleaned and converted to machine understandable format.

Types of NoSQL:

• Column
• Document
• Key-Value
• Graph
• Multi-model

To learn more about NoSQL and its types, refer here.

For example, a column called “name of cities” with values such as Delhi, Mumbai, Chennai, etc.

We can see that there is no order between the variables – viz Delhi is in no particular way higher or lower than Mumbai (unless explicitly mentioned).

The normal distribution is the most important and most widely used distribution in statistics. It is sometimes called the bell curve, because it has a peculiar shape of a bell. Mostly, a binomial distribution is similar to normal distribution. The difference between the two is normal distribution is continuous.

For example, if you have a variable ranging from 0 to 1 and other from 0 to 1000, you can normalize the variable, such that both are in the range 0 to 1.

• a powerful N-dimensional array object
• sophisticated (broadcasting) functions
• tools for integrating C/C++ and Fortran code
• useful linear algebra, Fourier transform, and random number capabilities

Besides its obvious scientific uses, NumPy can also be used as an efficient multi-dimensional container of generic data. Arbitrary data-types can be defined. This allows NumPy to seamlessly and speedily integrate with a wide variety of databases.

Word

Description

For example, if the data is

This is converted as

It consists of two parts:

1. Workflow engine: Responsibility of a workflow engine is to store and run workflows composed of Hadoop jobs e.g., MapReduce, Pig, Hive.
   
2. Coordinator engine: It runs workflow jobs based on predefined schedules and availability of data.

Features of Oozie:

• Oozie has client API and command line interface which can be used to launch, control and monitor job from Java application.
• Using its Web Service APIs one can control jobs from anywhere.
• Oozie has provision to execute jobs which are scheduled to run periodically.

• Reduce the model complexity
• Regularization

Word

Description

• A fast and efficient DataFrame object for data manipulation with integrated indexing.
• Tools for reading and writing data between in-memory data structures and different formats: CSV and text files, Microsoft Excel, SQL databases, and the fast HDF5 format.
• Flexible reshaping and pivoting of data sets.
• Columns can be inserted and deleted from data structures for size mutability.
• High performance merging and joining of data sets.

For more information on Pandas, you can refer here.

For instance, the weights in linear and logistic regression fall under the category of parameters.

In computer vision, supervised pattern recognition techniques are used for optical character recognition (OCR), face detection, face recognition, object detection, and object classification.

This represents a pie graph showing the results of an exam. Each grade is denoted by a “slice”. The total of the percentages is equal to 100. The total of the arc measures is equal to 360 degrees. So 12% students got A grade, 29% got B, and so on.

Key features of Pig:

• It is able to store data at any point during a pipeline.
• It declares execution plans.
• Supports pipeline splits, thus allowing workflows to proceed along DAGs instead of strictly sequential pipelines.
• Users can create their own functions to do special-purpose processing.

To read further on Pig, refer here.

For example, if you want to build a self learning car. You can spend years to build a decent image recognition algorithm from scratch or you can take inception model (a pre-trained model) from Google which was built on ImageNet data to identify images in those pictures.

For more information regarding Pre-trained model and their usages, you can refer here.

It can be represented as:

Precision = TP / (TP + FP)

Whereas recall is described as the measured of how many of the positive predictions were correct.

It can be represented as:

Recall = TP / (TP + FN)

It is a statistical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables called principal components.

PCA is mostly used as a tool in exploratory data analysis and for making predictive models. It’s often used to visualize genetic distance and relatedness between populations.

• Easy to learn.
• High-level language
• Broadly used and supported

To learn python from scratch, you can follow this article.

• Easy to use API
• Python support
• Dynamic computation graphs

Word

Description

Quartile divides a series into 4 equal parts. For any series, there are 4 quartiles denoted by Q1, Q2, Q3 and Q4. These are known as First Quartile , Second Quartile and so on.

For example, the diagram below shows the health score of a patient from range 0 to 60. Quartiles divide the population into 4 groups.

Word

Description

Features of R:

1. It is platform independent, so it is compatible with multiple operating systems
2. R has a very strong and consistent online community support
3. The graphical capabilities of R are awesome
4. There is abundance of literature to learn R

Suppose we have a dataset having 10 data points, listed below:

4,5,2,8,4,7,6,4,6,3

So, first of all we will arrange these data points in ascending order:

2,3,4,4,4,5,6,6,7,8

Now the range of this set is the difference between the highest(8) and the lowest(2) value.

Range = 8-2 = 6

• Content based filtering
• Collaborative filtering
  • User-User collaborative filtering
  • Item-Item collaborative filtering
• Hybrid recommendation systems

It is supervised learning method where the output variable is a real value, such as “amount” or “weight”.

Example of Regression: Linear Regression, Ridge Regression, Lasso Regression

Read more about this topic here.

 It is an example of machine learning where the machine is trained to take specific decisions based on the business requirement with the sole motto to maximize efficiency (performance). The idea involved in reinforcement learning is: The machine/ software agent trains itself on a continual basis based on the environment it is exposed to, and applies it’s enriched knowledge to solve business problems. This continual learning process ensures less involvement of human expertise which in turn saves a lot of time!

Important Note: There is a subtle difference between Supervised Learning and Reinforcement Learning (RL). RL essentially involves learning by interacting with an environment. An RL agent learns from its past experience, rather from its continual trial and error learning process as against supervised learning where an external supervisor  provides examples.

A good example to understand the difference is self driving cars. Self driving cars use Reinforcement learning to make decisions continuously like which route to take, what speed to drive on, are some of the questions which are decided after interacting with the environment. A simple manifestation for supervised learning would be to predict the total fare of a cab at the end of a journey.

Ridge regression performs ‘L2 regularization‘, i.e. it adds a factor of sum of squares of coefficients in the optimization objective. Thus, ridge regression optimizes the following:

Objective = RSS + α * (sum of square of coefficients)

Here, α (alpha) is the parameter which balances the amount of emphasis given to minimizing RSS vs minimizing sum of squares of coefficients. α can take various values:

1. α = 0:
   • The objective becomes same as simple linear regression.
   • We’ll get the same coefficients as simple linear regression.
2. α = ∞:
   • The coefficients will be zero. This is because of infinite weightage on square of coefficients, anything less than zero will make the objective infinite.
3. 0 < α < ∞:
   • The magnitude of α will decide the weightage given to different parts of objective.
   • The coefficients will be somewhere between 0 and 1 for simple linear regression.

Hence, for each sensitivity, we get a different specificity. The two vary as follows:

The ROC curve is the plot between sensitivity and (1- specificity). (1- specificity) is also known as false positive rate and sensitivity is also known as True Positive rate. Following is the ROC curve for the case in hand.

Let’s take an example of threshold = 0.5 (refer to confusion matrix). Here is the confusion matrix :

As you can see, the sensitivity at this threshold is 99.6% and the (1-specificity) is ~60%. This coordinate becomes on point in our ROC curve. To bring this curve down to a single number, we find the area under this curve (AUC).

Note that the area of entire square is 1*1 = 1. Hence, AUC itself is the ratio under the curve and the total area.

Here,

• Predicted -> value predicted by the model
• Actual -> observed values
• N -> Total number of observations

is invariant under rotations of the plane around the origin, because for a rotated set of coordinates through any angle θ.

Word

Description

• Its an object-oriented language that supports many traditional design patterns
• It supports functional programming which enables it to handle distributed programming at fundamental level
• It is designed to run on JVM platform that helps in directly using Java libraries
• Scala can be easily implemented into existing java projects as Scala libraries can be used within Java code
• It supports first-class objects and anonymous functions

These problems sit in between both supervised and unsupervised learning.

A good example is a photo archive where only some of the images are labeled, (e.g. dog, cat, person) and the majority are unlabeled.

Skewness is a measure of symmetry. A distribution, or data set, is symmetric if it looks the same to the left and right of the center point.

1. It connects to mathematics at all levels, from kindergarten to calculus
2. It is innate in humans
3. Spatial-temporal reasoning abilities can be increased. This understanding of Spatial-temporal reasoning forms the foundation of Spatial-temporal Math

The learning can be much faster with stochastic gradient descent for very large training datasets and often one only need a small number of passes through the dataset to reach a good or good enough set of coefficients.

Here, The goal is to approximate the mapping function so well that when you have new input data (x) that you can predict the output variables (Y) for that data.

Examples of Supervised Learning algorithms: Regression, Decision Tree, Random Forest, KNN, Logistic Regression etc.

It is a classification method. In this algorithm, we plot each data item as a point in n-dimensional space (where n is the number of features you have) with the value of each feature being the value of a particular coordinate.

For example, if we only have two features like Height and Hair length of an individual, we’d first plot these two variables in two-dimensional space where each point has two coordinates (these coordinates are known as Support Vectors) Now, we will find some line that splits the data between the two differently classified groups of data. This will be the line such that the distances from the closest point in each of the two groups will be farthest away.

Word

Description

Here you can check some of the most widely used pre-trained models.

Word

Description

Word

Description

Variance is used to measure the spread of given set of numbers and calculated by the average of squared distances from the mean

Let’s take an example, suppose the set of numbers we have is (600, 470, 170, 430, 300)
To Calculate:
1) Find the Mean of set of numbers, which is (600 + 470 + 170 + 430 + 300) / 5 = 394
2) Subtract the mean from each value which is (206, 76, -334, 36, -94)
3) Square each deviation from the mean which is (42436, 5776, 50176, 1296, 8836)
4) Find the Sum of Squares which is 108520
5) Divide by total number of items (numbers) which is 21704

Word

Description

Principal at a certain school claims that the students in his school are above average intelligence. A random sample of thirty students has a mean IQ score of 112. The mean population IQ is 100 with a standard deviation of 15. Is there sufficient evidence to support the principal’s claim?

So we can make use of z-test to test the claims made by the principal. Steps to perform z-test:

• Stating null hypothesis and alternate hypothesis.
• State the alpha level. If you don’t have an alpha level, use 5% (0.05).
• Find the rejection region area (given by your alpha level above) from the z-table. An area of .05 is equal to a z-score of 1.645.
• Find the test statistics using this formula:

Here,

• x ̅is the sample mean
• σ is population standard deviation
• n is sample size
• μ is the population mean

If the test statistic is greater than the z-score of rejection area, reject the null hypothesis. If it’s less than that z-score, you cannot reject the null hypothesis.

To get a better understanding of the topic, refer here.