All You Need to Know About Neural Networks

a high-level algorithms expert will provide a brief overview of the evolution of neural networks and discuss the latest approaches in the field.

Neural networks and deep learning technologies underpin most of the advanced intelligent applications today. In this article, Dr. Sun Fei (Danfeng), a high-level algorithms expert from Alibaba's Search Department, will provide a brief overview of the evolution of neural networks and discuss the latest approaches in the field. The article is primarily centered on the following five items:

• The Evolution of Neural Networks
• Sensor Models
• Feed-forward Neural Networks
• Back-propagation
• Deep Learning Basics

1. The Evolution of Neural Networks

Before we dive into the historical development of neural networks, let's first introduce the concept of a neural network. A neural network is primarily a computing model that simulates the workings of the human brain at a simplified level. This type of model uses a large number of computational neurons which connect via layers of weighted connections. Each layer of neurons is capable of performing large-scale parallel computing and passing information between them.

The timeline below shows the evolution of neural networks:

The origin of neural networks goes back to even before the development of computing itself, with the first neural networks appearing in the 1940s. We will go through a bit of history to help everyone gain a better understanding of the basics of neural networks.

The first generation of neural network neurons worked as verifiers. The designers of these neurons just wanted to confirm that they could build neural networks for computation. These networks cannot be used for training or learning; they simply acted as logic gate circuits. Their input and output was binary and weights were predefined.

The second phase of neural network development came about in the 1950s and 1960s. This involved Roseblatt's seminal work on sensor models and Herbert's work on learning principles.

2. Sensor Models

Sensor models and the neuron models as we mentioned above were similar but had some key differences. The activation algorithm in a sensor model can be either a break algorithm or a sigmoid algorithm, and its input can be a real number vector instead of the binary vectors used by the neuron model. Unlike the neuron model, the sensor model is capable of learning. Next, we will talk about some of the special characteristics of the sensor model.

We can think of the input value (x1..., xn) as a coordinate in N dimensional space, while wTx-w0 = 0 is a hyperplane in N dimensional space. Obviously, if wTx-w0 < 0, then the point falls below the hyperplane, while if wTx-w0 > 0, then the point falls above the hyperplane.

The sensor model corresponds to the hyperplane of a classifier and is capable of separating different types of points in N dimensional space. Looking at the figure below, we can see that the sensor model is a linear classifier.

The sensor model is capable of easily performing classification for basic logical operations like AND, OR, and NOT.

Can we classify all logical operations through the sensor model? The answer is, of course not. For example, Exclusive OR operations are very difficult to classify through a single linear sensor model, which is one of the main reasons that neural networks quickly entered a low point in development soon after the first peak. Several authors, including Minsky, discussed this problem on the topic of sensor models. However, a lot of people misunderstood the authors on this subject.

In reality, authors like Minsky pointed out that one could implement Exclusive OR operations through multiple layers of sensor models; however, since the academic world lacked effective methods to study multi-layer sensor models at the time, the development of neural networks dipped into its first low point.

The figure below shows intuitively how multiple layers of sensor models can achieve Exclusive OR operations:

3. Feed-Forward Neural Networks

Entering into the 1980s, due to the expressive ability of sensor model neural networks being limited to linear classification tasks, the development of neural networks began to enter the phase of multi-layered sensors. A classic multi-layered neural network is a feed-forward neural network.

We can see from the figure below that it involves an input layer, a hidden layer with an undefined number of nodes, and an output layer.

We can express any logical operation by a multi-layer sensor model, but this introduces the issue of weighted learning between the three layers. When xk is transferred from the input layer to the weighted vkj on the hidden layer and then passed through an activation algorithm like sigmoid, then we can retrieve the corresponding value hj from the hidden layer. Likewise, we can use a similar operation to derive the yi node value from the output layer using the hj value. For learning, we need the weighted information from the w and v matrices so that we can finally obtain the estimated value y and the actual value d.

If you have a basic understanding of machine learning, you will understand why we use a gradient descent to learn a model. The principle behind applying a gradient descent to the sensor model is fairly simple, as we can see from the below figure. First, we have to determine the model's loss.

The example uses a square root loss and seeks to close the gap between the simulated value y and the real value d. For the sake of convenient computing, in most situations, we use the root relationship E = 1/2 (d-y)^2 = 1/2 (d-f(x))^2.

According to the gradient descent principle, the rate of the weighting update cycle is: wj ← wi + α(d − f(x))f′(x)xi, where α is the rate of learning which we can adjust manually.

4. Back-Propagation

How do we learn all of the parameters in a multilayer feed-forward neural network? The parameters for the top layer are very easy to obtain. One can achieve the parameters by comparing the difference between the estimated and real values output by the computing model and using the gradient descent principles to obtain the parameter results. The problem comes when we try to obtain parameters from the hidden layer. Even though we can compute the output from the model, we have no way of knowing what the expected value is, so we have no way of effectively training a multi-layer neural network. This issue plagued researchers for a long time, leading to the lack of development of neural networks after the 1960s.

Later, in the 70s, a number of scientists independently introduced the idea of a back-propagation algorithm. The basic idea behind this type of algorithm is actually quite simple. Even though at the time there was no way to update according to the expected value from the hidden layer, one could update the weights between the hidden and other layers via errors passed from the hidden layer. When computing a gradient, since all of the nodes in the hidden layer are related to multiple nodes on the output layer, so all of the layers on the previous layers are accumulated and processed together.

Another advantage of back-propagation is that we can perform gradients and weighting of nodes on the same layer at the same time since they are unrelated. We can express the entire process of back-propagation in pseudocode as below:

Next, let's talk about some of the other characteristics of a back-propagation neural network. A back-propagation is actually a chain rule. It can easily generalize any computation that has a map. According to the gradient function, we can use a back-propagation neural network to produce a local optimized solution, but not a global optimized solution. However, from a general perspective, the result produced by a back-propagation algorithm is usually a satisfactorily optimized solution. The figure below is an intuitive representation of a back-propagation algorithm:

Under most circumstances, a back-propagation neural network will find the smallest possible value within scope; however, if we leave that scope we may find an even better value. In actual application, there are a number of simple and effective ways to address this kind of issue, for example, we can try different randomized initialization methods. Moreover, in practice, among the models frequently used in the field of modern deep learning, the method of initialization has significant influence on the final result. Another method of forcing the model to leave the optimized scope is to introduce random noises during training or use a hereditary algorithm to prevent the training model from stopping at an un-ideal optimized position.

A back-propagation neural network is an excellent model of machine learning, and when speaking about machine learning, we can't help but notice a basic issue that's frequently encountered throughout the process of machine learning, that is the issue of overfitting. A common manifestation of overfitting is that during training, even when the loss of the model constantly drops, the loss and error in the test group rises. There are two typical methods to avoid overfitting:

• Early stopping: we can separate a validation group ahead of time, and run it against this already verified group during training. We can then observe the loss of the model and, if the loss has already stopped dropping in the verification group but is still dropping in the training group, then we can stop the training early to prevent overfitting.
• Regularization: we can add rules to the weights within the neural network. The dropout method, which is popular these days, involves randomly dropping some nodes or sides. This method, which we can consider as a form of regularization, is extremely effective at preventing overfitting.

Even though neural networks were very popular during the 1980s, they, unfortunately, entered another low point in development in the 1990s. A number of factors contributed to this low point. For example, the Support Vector Machines, which were a popular model in the 1990s, took the stage at all kinds of major conferences and found application in a variety of fields. Support Vector Machines have an excellent statistical learning theory and are easy to intuitively understand. They are also very effective and produce near-ideal results.

Amidst this shift, the rise of the statistical learning theory behind Support Vector Machines applied no small amount of pressure to the development of neural networks. On the other hand, from the perspective of neural networks themselves, even though you can use back-propagation networks to train any neural network in theory, in actual application, we notice that as the number of layers in the neural network increases, the difficulty of training the network increases exponentially. For example, in the beginning of the 1990s, people noticed that in a neural network with a relatively large number of layers, it was common to see gradient loss or gradient explosion.

A simple example of gradient loss, for example, would be where each layer in a neural network is a sigmoid structure layer, therefore its loss during back-propagation is chained into a sigmoid gradient. When a series of elements are strung together, then if one of the gradients is very small, the gradient will then become smaller and smaller. In reality, after propagating one or two layers, this gradient disappears. Gradient loss leads to the parameters in deep layers to stop changing, making it very difficult to get meaningful results. This is one of the reasons that a multi-layer neural network can be very difficult to train.

The academic world has studied this issue in-depth, and come to the conclusion that the easiest way to handle it is by changing the activation algorithm. In the beginning, we tried to use a Rectified activation algorithm since the sigmoid algorithm is an index method which can easily bring about the issue of gradient loss. Rectified, on the other hand, replaces the sigmoid function and replaces max (0,x). From the figure below we can see that the gradient for estimates above 0 is 1, which prevents the issue of gradient disappearance. However, when the estimate is lower than 0, we can see that the gradient is 0 again, so the ReLU algorithm must be imperfect. Later, a number of improved algorithms came out, including Leaky ReLU and Parametric Rectifier (PReLU). When the estimate x is smaller than 0, we can convert it to a coefficient like 0.01 or α to prevent it from actually being 0.

With the development of neural networks, we later came up with a number of methods that solve the issue of passing gradients on a structural level. For example, the Metamodel, LSTM model, and modern image analysis use a number of cross-layer linking methods to more easily propagate gradients.

5. Deep Learning Basics

From the second low point in development in the 1990s to 2006, neural networks once again entered the consciousness of the masses, this time in even more force than before. A monumental occurrence during this rise of neural networks was the two theses on multi-layer neural networks (now called “deep learning”) submitted by Hinton in a number of places including Salahundinov.

One of these theses solved the issue of setting initialization values for neural networks. The solution, put simply, is to consider the input value as x, and the output value as decoded x, then through this method find a better initialization point. The other thesis raised a method of quickly training a deep neural network. Actually, there are a number of factors contributing to the modern popularity of neural networks, for example, the enormous growth in computing resources and availability of data. In the 1980s, it was very difficult to train a large scale neural network due to the lack of data and computing resources.

The early rise of neural networks was driven by three monumental figures, namely Hinton, Bengio, and LeCun. Hinton's main accomplishment was in the Restricted Boltzmann Machine and Deep Autoencoder. Bengio's major contribution was a series of breakthroughs in using the metamodel for deep learning. This was also the first field in which deep learning experienced a major breakthrough.

In 2013, language modeling, based on the metamodel, was already capable of outperforming even the most effective method at the time, the probability model. The main accomplishment of LeCun was research related to CNN. The primary appearance of deep learning was in a number of major summits like NIPS, ICML, CVPR, ACL, where it attracted no small amount of attention. This included the appearance of Google Brain, Deep Mind, and Facebook AI, which all placed the center of their research on the field of deep learning.

The first breakthrough to come about after deep learning entered the consciousness of the masses was in the field of speech recognition. Before we began using deep learning, models were all trained on previously defined statistical databases. In 2010, Microsoft used a deep learning neural network for speech recognition. We can see from the figure below that two error indicators both dropped by 2/3, an obvious improvement. Based on the newest ResNet technology, Microsoft has already reduced this indicator to 6.9%, with improvements coming year by year.

In the field of image classification, the CNN model experienced a major breakthrough in the form of ImageNet in 2012. In ImageNet, the Image classification is tested using a massive data collection and then sorted into 1000 types. Before the application of deep learning, the best error rate for image classification system was 25.8% (in 2011), which came down to a mere 10%, thanks to the work done by Hinton and his students in 2012 using CNN.

From the graph, we can see that since 2012, this indicator has experienced a major breakthrough each year, all of which have been achieved using the CNN model.
These massive achievements owe in large part to the multi-layered structure of modern systems, as they allow for independent learning and the ability to express data through a layered abstraction structure. The abstracted features can be applied to a variety of tasks, contributing significantly to the current popularity of deep learning.

Next, we will introduce two classic and common types of deep learning neural networks: One is the Convolutional Neural Network (CNN), and the other is the Recurrent Neural Network

Convolutional Neural Networks

There are two core concepts to Convolutional Neural Networks. One is convolution and the other is pooling. At this point, some may ask why we don't simply use feed-forward neural networks rather than CNN. Taking a 1000x1000 image, for example, a neural network would have 1 million nodes on the hidden layer. A feed-forward neural network, then, would have 10^12 parameters. At this point, it's nearly impossible for the system to learn since it would require an absolutely massive number of estimations.

However, a large number of images have characteristics like this. If we use CNN to classify images, then because of the concept of convolution, each node on the hidden layer only needs to connect and scan the features of one location of the image. If each node on the hidden layer connects to 10*10 estimations, then the final number of parameters is 100 million, and if the local parameters accessed by multiple hidden layers can be shared, then the number of parameters is decreased significantly.

Looking at the image below, the difference between feed-forward neural networks and CNN is obviously massive. The models in the image are, from left to right, fully connected, normal, feed-forward, fully connected feed-forward, and CNN modeled neural networks. We can see that the connection weight parameters of nodes on the hidden layer of a CNN neural network can be shared.

Another operation is pooling. A CNN will, on the foundation of the principle of convolution, form a hidden layer in the middle, namely the pooling layer. The most common pooling method is Max Pooling, wherein nodes on the hidden layer choose the largest output value. Because multiple kernels are pooling, we get multiple hidden layer nodes in the middle.

What is the benefit? First of all, pooling further reduces the number of parameters, and secondly, it provides a certain amount of translation invariance. As shown in the image, if one of the nine nodes shown in the image were to experience translation, then the node produced on the pooling layer would remain unchanged.

These two characteristics of CNN have made it popular in the field of image processing, and it has become a standard in the field of image processing. The example of the visualized car below is a great example of the application of CNN in the field of image classification. After entering the original image of the car into the CNN model, we can pass some simple and rough features like edges and points through the convolution and ReLU activation layer. We can intuitively see that the closer they are to the output image from the uppermost output layer, the closer they are to the contours of a car. This process will finally retrieve a hidden layer representation and connect it to the classification layer, after which it will receive a classification for the image, like the car, truck, airplane, ship, and horse shown in the image.

The image below is a neural network used in the early days by LeCun and other researchers in the field of handwriting recognition. This network found application in the US postal system in the 1990s. Interested readers can log into LeCun's website to see the dynamic process of handwriting recognition.

While CNN has become incredibly popular in the field of image recognition, it has also become instrumental in text recognition over the past two years. For example, CNN is currently the basis of the most optimal solution for text classification. In terms of determining the class of a piece of text, all one really needs to do is look for indications from keywords in the text, which is a task that is well suited to the CNN model.

CNN has widespread real-world applications, for example in investigations, self-driving cars, Segmentation, and Neural Style. Neural Style is a fascinating application. For example, there is a popular app in the App Store called Prisma, which allows users to upload an image and convert it into a different style. For example, it can be converted to the style of Van Goh's Starry Night. This process relies heavily on CNN.

Recursive Neural Networks

As for the foundational principles behind recursive neural networks, we can see from the image below that the output from such a network relies not only on output x but the status of the hidden layer, which is updated according to the previous input x. The expanded image shows the entire process. The hidden layer from the first input is S(t-1), which influences the next input, X(t). The main advantage of the recursive neural network model is that we can use it in sequential data operations like text, language, and speech where the state of the current data is influenced by previous data states. This type of data is very difficult to handle using a feed-forward neural network.

Speaking of recursive neural networks, we would be remiss not to bring up the LSTM model we mentioned earlier. LSTM is not actually a complete neural network. Simply put, it is the result of an RNN node that has undergone complex processing. An LSTM has three gates, namely the input gate, the regret gate, and the output gate.
Each of these gates is used to process the data in a cell and determine whether or not the data in the cell should be input, regretted, or output.

Finally, let's talk a bit about a cross-discipline application of neural networks which is gaining widespread acceptance. This application involves converting an image into a text description of the image or a title describing it. We can describe the specific implementation process by using a CNN model first to extract information about the image and produce a vector representation. Later on, we can pass that vector as input to an already trained recursive neural network to produce the description of the image.

Summary

In this article, we talked about the evolution of neural networks and introduced several basic concepts and approaches in this field.

Source: https://dzone.com/articles/all-you-need-to-know-about-neural-networks-part-2

How to apply Machine Learning and AI for Sales & Marketing

1. Clustering For Customer Segmentation And Discovery

Not all customers are the same. Unsupervised machine learning can help marketers group their audience into dynamic groups and engage them accordingly. For example, a system can analyze billions of consumer interest variables, identifies specific customer’s interests based on their social media activities, then generates a visual report grouping people with similar interests. You then gain insight on which of your customers are die-hard foodies, who follows which series on Netflix, or who among them have similar travel plans.

2. Multi-Arm Contextual Bandits For Content Optimization

A/B tests are effective ways of finding out which content option (email tone, web page layout, visual elements in an ad, article headline, etc.) resonates better with your audience. However, A/B Testing involves a period of “regret” where you lose revenue while using the less optimal option. You have to wait and finish the countdown before learning which option — the final answer — is better. In contrast, bandit tests mitigate regret (opportunity loss) through dynamic optimization where it simultaneously explores and exploits options, gradually and automatically moving towards the better option.

 

3. Regression Models For Dynamic Pricing

 The right pricing scheme can make or break a product. Regression techniques in machine learning allow marketers to predict numerical values based on pre-existing features, which in turn enables them to optimize different aspects of the customer journey. Regression can also be used in sales forecasting and in optimizing marketing spend.

Regression Models for Price Prediction

4. Text Classification For User Insight And Personalization

Using natural language processing (NLP), a machine learning system can probe text- or voice-based content, then classify each content based on variables such as tone, sentiment, or topic to generate consumer insight or curate relevant materials. IBM Watson’s Tone Analyzer, for example, can parse through online customer feedback and determine the general tone of users reviewing a product.

5. Text Extraction And Summarization For Trending News

Social News Analytics & Summarization

  

Marketers can leverage ML to extract relevant content from online news articles and other data sources to determine how people view their brand and/or react to their products

 6. Attentional Neural Networks For Machine Translation

Attention mechanisms in deep learning help improve machine translation and empower your marketing assets for the global stage. Translation work for a brand’s entry into a new, linguistically different market used to be a major marketing spend but advances in AI enable machine translation to achieve near human parity. To rationalize costs and speed up the process, many companies opt to just have a human translator review and sign off machine translation output.

7. Recurrent Neural Networks (RNN) For Text Generation

http://redcatlabs.com/2017-06-22_TFandDL_Captioning/#/

If your branding creatives are constantly pressured to come up with great names for new products, campaigns, and companies, you can use generative models like RNNs to serve up loads of plausible-sounding names — some catchy, some weird, and a few surprisingly on the spot. 

8. Dialog Systems For Chatbots And Customer Experience Automation

Bots and chatbots represent one of the most ubiquitous applications of AI, but most marketing bots you see in the wild are completely scripted and use minimal natural language processing and machine learning. The more sophisticated dialog systems are able to reference external knowledge bases, adapt to unusual questions, and also escalate to human agents when required. Quite a number of companies have already adopted chatbots to engage customers throughout their lifecycle, from when they first learn of a brand to after they've already made purchases and require customer support.

9. Text-To-Speech (TTS) And Speech-To-Text (STT) To Power Voice-Based Search

Considered part of the conversational AI domain, voice-enabled and voice-only platforms introduce a new paradigm and new user engagement possibilities into our software and hardware interfaces. With the rising adoption of voice-based digital assistants such as Amazon Echo and Google Assistant that enable touch-free shopping and search, marketing executives need a conversational AI strategy to future-proof their marketing.

10. Computer Vision For Branded Object Recognition

 Computer vision is a rapidly advancing field in AI that lends itself to a wide range of applications. Marketers can use ML-powered computer vision for product recognition and to extract user insight from unlabeled images and videos. Solutions like GumGum allow marketers to identify when their brand logos have appeared in user-generated content and quickly calculate earned media from video analysis. More tech-savvy marketers can use an API like Clarifai to build custom solutions for content moderation as well as search and recommendation engines based on visual similarity.

12. Automated Data Visualization For Superior Reporting

Images speak louder than words. AI is a lot faster and more efficient at transforming data into visual insight than any human expert. Analysts usually use tools like Excel or Tableau to manually create visualizations, but automated enterprise analytics solutions such as Qlik can centralize data sources and generate useful dashboards and reports for your marketing teams. Many platforms now use data analytics and advanced machine learning algorithms to vividly clarify market trends, people’s behavioral patterns, and other information that are otherwise hidden from plain view and not readily convertible to practical insight.

13. Reinforcement Learning For Sequential Marketing Decisions

Reinforcement Learning

 Some of the most complex decisions we make are not single predictions, but rather a series of decisions made over a long time horizon. Balancing short-term tradeoffs versus long-term gains is challenging even for the smartest humans.

Reinforcement learning has been used successfully in cases like DeepMind's AlphaGo to beat human decision making in such complex scenarios. While business cases are usually far more complex than games, the success in narrow domains suggests promise for larger ones. A notable study by IBM researchers explores how reinforcement learning could be used to optimize targeted marketing.

Data Science Interview Questions

If you're looking for Data Science Interview Questions & Answers for Experienced or Freshers, you are at right place. There are lot of opportunities from many reputed companies in the world. According to research Data Science Market Expected to Reach $128.21 Billion With 36.5% CAGR Forecast To 2022. So, You still have opportunity to move ahead in your career in Data Science Analytics. 

Q. What Is Data Science?
Data Science is a new area of specialization being developed by the Department of Mathematics and Statistics at Bowling Green State University. This field integrates math, statistics, and computer science to prepare for the rapidly expanding need for data scientists. Students seeking to pursue studies in Data Science should declare a mathematics major as entering freshmen, in anticipation of completing the specialization in years three and four.

Q. What are the requirements of a Data Science program?
The Data Science specialization requires three semesters of calculus (MATH 1310, MATH 2320, and MATH 2330 or MATH 2350), linear algebra (MATH 3320), introduction to programming (CS 2010), probability and statistics I (MATH 4410), and regression analysis (STAT 4020). In addition, the requirements include:
a) MATH 2950 Introduction to Data Science. This one-hour seminar would introduce freshmen students to a variety of data-science applications and give them an introduction to programming.
b) MATH 3430 Computing with Data. This course will focus on the data wrangling and data exploration computational skills in the context of a modern computing language such as Python or R.
c) MATH 3440 Statistical Programming. This course will focus on writing scripts and functions using a modern statistical language such as R.
d) MATH 4440 Statistical Learning. This course deals with modern methods for modeling data including a variety of supervised and unsupervised methods.
In addition, the student will be asked to choose two of the following seven courses:
a) MATH 4320 Linear Algebra with Applications
b) MATH 4420 Probability and Statistics II
c) MATH 4470 Exploratory Data Analysis
d) CS 2020 Object-Oriented Programming
e) STAT 4440 Data Mining in Business Analytics
f) CS 4400 Optimization Techniques
g) CS 4620 Database Management Systems

Q. Compare SAS, R, Python, Perl?
a) SAS is a commercial software. It is expensive and still beyond reach for most of the professionals (in individual capacity). However, it holds the highest market share in Private Organizations. So, until and unless you are in an Organization which has invested in SAS, it might be difficult to access one. R & Python, on the other hand are free and can be downloaded by any one.
b) SAS is easy to learn and provides easy option (PROC SQL) for people who already know SQL. Even otherwise, it has a good stable GUI interface in its repository. In terms of resources, there are tutorials available on websites of various university and SAS has a comprehensive documentation. There are certifications from SAS training institutes, but they again come at a cost. R has the steepest learning curve among the 3 languages listed here. It requires you to learn and understand coding. R is a low level programming language and hence simple procedures can take longer codes. Python is known for its simplicity in programming world. This remains true for data analysis as well.
c) SAS has decent functional graphical capabilities. However, it is just functional. Any customization on plots are difficult and requires you to understand intricacies of SAS Graph package. R has the most advanced graphical capabilities among the three. There are numerous packages which provide you advanced graphical capabilities. Python capabilities will lie somewhere in between, with options to use native libraries (matplotlib) or derived libraries (allowing calling R functions).
d) All 3 ecosystems have all the basic and most needed functions available. This feature only matters if you are working on latest technologies and algorithms. Due to their open nature, R & Python get latest features quickly (R more so compared to Python). SAS, on the other hand updates its capabilities in new version roll-outs. Since R has been used widely in academics in past, development of new techniques is fast.

Q. Mention features of Teradata?
a) Parallel architecture:The Teradata Database provides exceptional performance using parallelism to achieve a single answer faster than a non-parallel system. Parallelism uses multiple processors working together to accomplish a task quickly.
b) Single datastore:The Teradata Database acts as a single data store, instead of replicating database for different purposes with the teradata database we can store the data once and use it for all applications. The Teradata database provides same connectivity for all systems.
c) Scalability:Scalability is nothing but we can add components to the system, the performance increase as linear. Scalability enables the system to grow to support more users/data/queries/complexity of queries without experiencing performance degradation.

Q. What does a data scientist do?
A data scientist represents an evolution from the business or data analyst role. The formal training is similar, with a solid foundation typically in computer science and applications, modeling, statistics, analytics and math. What sets the data scientist apart is strong business acumen, coupled with the ability to communicate findings to both business and IT leaders in a way that can influence how an organization approaches a business challenge. Good data scientists will not just address business problems, they will pick the right problems that have the most value to the organization. A traditional data analyst may look only at data from a single source – a CRM system, for example – a data scientist will most likely explore and examine data from multiple disparate sources. The data scientist will sift through all incoming data with the goal of discovering a previously hidden insight, which in turn can provide a competitive advantage or address a pressing business problem. A data scientist does not simply collect and report on data, but also looks at it from many angles, determines what it means, then recommends ways to apply the data.

Q. How do Data Scientists Code in R?
R is a popular open source programming environment for statistics and data mining. The good news is that it is easily integrated into ML Studio. I have a lot of friends using functional languages for machine learning, such as F#. It’s pretty clear, however, that R is dominant in this space. Polls and surveys of data miners are showing R’s popularity has increased substantially in recent years. R was created by Ross Ihaka and Robert Gentleman at the University of Auckland, New Zealand, and is currently developed by the R Development Core Team, of which Chambers is a member. R is named partly after the first names of the first two R authors. R is a GNU project and is written primarily in C, Fortran.

Q. What is Machine Learning?
Machine learning represents the logical extension of simple data retrieval and storage. It is about developing building blocks that make computers learn and behave more intelligently. Machine learning makes it possible to mine historical data and make predictions about future trends. Search engine results, online recommendations, ad targeting, fraud detection, and spam filtering are all examples of what is possible with machine learning. Machine learning is about making data-driven decisions. While instinct might be important, it is difficult to beat empirical data.

Q. What is the use of Machine Learning?
Machine Learning is found in things we use every day such as Internet search engines, email and online music and book recommendation systems. Credit card companies use machine learning to protect against fraud.
Using adaptive technology, computers recognize patterns and anticipate actions. Machine Learning is used in more complex applications such as:
1. Self-parking cars
2. Guiding robots
3. Airplane navigation systems (manned and unmanned),
4. Space exploration
5. Medicine

Q. What is Machine Learning best suited for?
Machine Learning is good at replacing labor-intensive decision-making systems that are predicated on hand-coded decision rules or manual analysis. Six types of analysis that Marchine Learning is well suited for are:
1. classification (predicting the class/group membership of items)
2. regression (predicting real-valued attributes)
3. clustering (finding natural groupings in data)
4. multi-label classification (tagging items with labels)
5. recommendation engines (connecting users to items)

Q. Define Boxplot?
In descriptive statistics, a boxplot, also known as a box-and-whisker diagram or plot, is a convenient way of graphically depicting groups of numerical data through their FIVE-NUMBER SUMMARIES (the smallest observation, lower quartile (Q1), median (Q2), upper quartile (Q3), and largest observation). A boxplot may also indicate which observations, if any, might be considered outliers.

Q. What are the most important machine learning techniques?
In Associative rule learning computers are presented with a large set of observations, all being made up of multiple variables. The task is then to learn relations between variables such us A & B 3C (if A and B happen, then C will also happen).
In Clustering computers learn how to partition observations in various subsets, so that each partition will be made up of similar observations according to some well-defined metric. Algorithms like K-Means and DBSCAN belong also to this class.
In Density estimation computers learn how to find statistical values that describe data. Algorithms like Expectation Maximization belong also to this class.

Q. Why is it important to have a robust set of metrics for machine learning?
Any machine learning technique should be evaluated by using metrics for analytically assessing the quality of results. For instance: if we need to categorize objects such as people, movies or songs into different classes, precision and recall might be suitable metrics.
Precision is the ratio tp / tp+fp where tP is the number of true positives and fP is the number of false positives. Recall is the ratio tp / tP+fn where tP is the number of true positives and fn is the number of false negatives. True and false are attributes derived by using manually created data. Precision and Recall are typically reported in a 2-d graph known as P/R Curves, where different algorithmic graphs can be compared by reporting the achieved Precision for fixed values of Recall.
In addition F1 is another frequently used metric, which combines precision and Recall into a single value:
F1 = 2*precision*recall / precision+recall
Scikit-learn provides a comprehensive set of metrics for classification, clustering, regression, ranking and pairwise judgment’. As an example the code below computes Precision and Recall.
Code
import numpy as np
from skleam.metrics import precision_recall_curve
y_true = np.array([0,1,1,0, 1])
y_scores = np.array([0.5, 0.6, 0.38, 0.9, 1])
precision, recall, thresholds = precision_recall_curve(y_true, y_scores)
print precision
print recall

Q. Why are Features extraction and engineering so important in machine learning?
The Features are the selected variables for making predictions. For instance, suppose you’d like to forecast, if tomorrow there will be a sunny day, then you will probably pick features like humidity (a numerical value), speed of wind (another numeric value), some historical information (what happened during the last few years), whether or not it is sunny today (a categorical value yes/no) and a few other features. Your choice can dramatically impact on your model for the same algorithm and you need to run multiple experiments in order to find what the right amount of data and what the right features are in order to forecast with minimal error. It is not unusual to have problems represented by thousands of features and combinations of them and a good feature engineer will use tools for stack ranking features according to their contribution in reducing the error for prediction.
Different authors use different names for different features including attributes, variables and predictors. In this book we consistently use features.
Features can be categorical such as marital status, gender, state of residence, place of birth, or numerical such as age, income, height and weight. This distinction is important because certain algorithms such as linear regression work only with numerical attributes and if categorical features are present, they need to be somehow encoded into numerical values.
In other words, feature engineering is the art of extracting, selecting and transforming essential characteristics representing data. It is sometimes considered less glamourous than machine learning algorithms but in reality any experienced Data Scientist knows that a simple algorithm on a well-chosen set of features performs better than a sophisticated algorithm on a not so good set of features.
Also simple algorithms are frequently easier to implement in a distributed way and therefore they scale well with large datasets. So the rule of thumb is in what Galileo already said many centuries ago: “Simplicity is the ultimate sophistication”. Pick your algorithm carefully and spend a lot of time in investigating your data and in creating meaningful summaries with appropriate feature engineering.
Real world objects are complex and features are used to analytically represent those objects. From one hand this representation has an inherent error which can be reduced by carefully selecting a right set of representatives. From the other hand we might not want to create a too complex representation because it might be computationally expensive for the machine to learn a sophisticate model, indeed such model could possibly not generalize well to the unseen data.
Real world data is noisy. We might have very few instances (outliers) which show a sensible difference from the majority of the remaining data, while the selected algorithm should be resilient enough to outliers.
Real world data might have redundant information. When we extract features, we might be interested in optimizing simplicity of learned models and discard new features which show a high correlation with the already observed ones.
ETL is the process of Extraction, Transformation and Loading of features from real data for creating various learning sets. Transformation in particular refers to operations such as features weighting, high correlated features discarding, the creation of synthetic features derivative of the one observed in the data and the reduction of high dimension features space into a lower one by using either hashing or rather sophisticate space projection techniques. For example in this book we discuss:
1. TFxIDF, an example of features weighting used in text classification
2. ChiSquare, an example of filtering of highly correlated features
3. Kernel Trick, an example of creation of derivative features
4. Hashing, a simple technique to reduce feature space dimensions
5. Binning, an example of transformation of continuous features into a discrete one. New synthetic features might be created in order to represent the bins.

Q. Can you provide an example of features extraction?
Let’s suppose that we want to perform machine learning on textual files. The first step is to extract meaningful feature vectors from a text. A typical representation is the so called bag of words where:
1. Each word w in the text collection is associated with a unique integer = wordld(w) assigned to it.
2. For each document i, the number of occurrences of each word w is computed and this value is stored in a matrix M(i.D. Please, note that M is typically a sparse matrix because when a word is not present in a document, its count will be zero.
Numpy, Scikit-learn and Spark all support sparse vectors2. Let’s see an example where we start to load a dataset made up of Usenet articles; where the altatheism category is considered and the collection of text documents is converted into a matrix of token counts. We then print the wordId(‘man).
Code
From sklearn.datasets import fetch_20 new groups
HTTP://DOCS.SCIPY.ORG/DOC/SCIPY/REFERENCE/SPARSE.HTML
HTTP://SCIKIT-LEARN.ORG/STABLE/DATASETS/
Two additional observations can be here highlighted: first, the training set, the validation set and the test set are all sampled from the same gold set but those samples are independent. Second, it has been assumed that the learned model can be described by means of two different functions f and h combined by using a set of hyper-parameters A.
Unsupervised machine learning consists in tests and application phases only because there is no model to be learned a-priori. In fact unsupervised algorithms adapt dynamically to the observed data.

Q. What is a Bias – Variance tradeoff?
Bias and Variance are two independent sources of errors for machine learning which prevent algorithms to generalize the models learned beyond the training set.
a) Bias is the error representing missing relations between features and outputs. In machine learning this phenomenon is called underfitting.
b) Variance is the error representing sensitiveness to small training data fluctuations. In machine learning this phenomenon is called overfitting.
A good learning algorithm should capture patterns in the training data (low bias), but it should also generalize well with unseen application data. In general a complex model can show low bias because it captures many relations in the training data and, at the same time, it can show high variance because it will not necessarily generalize well. The opposite happens with models with high bias and low variance. In many algorithms an error can be analytically decomposed in three components: bias, variance and the irreducible error representing a lower bound on the expected error for unseen sample data.
One way to reduce the variance is to try to get more data or to decrease the complexity of a model. One way to reduce the bias is to add more features or to make the model more complex, as adding more data will not help in this case. Finding the right balance between Bias and Variance is an art that every Data scientist must be able to manage.

Q. What is a cross-validation and what is an overfitting?
Learning a model on a set of examples and testing it on the same set is a logical mistake because the model would have no errors on the test set but it will almost certainly have a poor performance on the real application data. This problem is called overfitting and it is the reason why the gold set is typically split into independent sets for training, validation and test. An example of random split is reported in the code section, where a toy dataset with diabetics’ data has been randomly split into two parts: the training set and the test set.
As discussed in the previous question: given a family of learned models, the validation set is used for estimating the best hyper-parameters. However by adopting this strategy there is still the risk that the hyper-parameters overlit a particular validation set.
The solution to this problem is called cross-validation. The idea is simple: the test set is split in k smaller sets called folds and the model is then learned on k – 1 folds, while the remaining data is used for validation. This process is repeated in a loop and the metrics achieved for each iterations are averaged. An example of cross validation is reported in the section below where our toy dataset is classified via SVM and accuracy is computed via cross-validation. 5VM is a classification technique and “accuracy” is a quality measurement, which will be discussed later in the book.
Stratified KFold is a variation of k-fold where each set contains approximately the same balanced percentage of samples for each target class as the complete set.
Code
import numpy as np
front skleam import cross_validation
from skleam import datasets
from skleam import svm
diabets = datasets.load_diabetes()
X_train, X_test, y_train, y_test =
cross_validation.train_test_split(
diabets.data, diabets.target, test_size*).2, random_state=0)
print X_train.shape, y_trainshape # test size 20%
print X_test.shape, y_test.shape
cif = svm.SVC(kemel=qineari, C=1)
scores = cross_validation.cross_valscore(
clf, diabets.data, diabets.target, cv=4) # 4-folds
print scores
print(“Accuracy: %0.2f (+/- %0.2f)” % (scores.meanQ, scores.std()))

Q. Why are vectors and norms used in machine learning?
Objects such as movies, songs and documents are typically represented by means of vectors of features. Those features are a synthetic summary of the most salient and discriminative objects characteristics. Given a collection of vectors (the so-called vector space) V, a norm on V is a function P: V A satisfying the following properties: For all complex numbers a and all
u, v E V,
1. P(av) = |a| p(v)
2. P(u+v) <= p(u) + p(v)
3. If p(v) = 0 then v is the zero vector
The intuitive notion of length for a vector is captured by
||x||2 = root (x12 + … + xn2 )
More generally we have

Code
from numpy import linalg as LA import numpy as np a = np.arange(22) print LA.nonn(a) print LA.nonn(a, I)

Q. What are Numpy, Scipy and Spark essential datatypes?
Numpy provides efficient support for memorizing vectors and matrices and for linear algebra operations 5. For instance: dot(a, b[, out]) is the dot product of two vectors, while inner(a, b) and outer(a, b[, out]) are respectively the inner and outer products.
Scipy provides support for sparse matrices and vectors with multiple memorization strategies in order to save space when dealing with zero entries.6 In particular the COOrdinate format specifies the non-zero v value for the coordinates(•), while the Compressed Sparse Colum matrix (CSC) satisfies the relationship M[row_ind[k], col_ind[k]] = data[k]
Spark has many native datatypes for local and distributed computations. The primary data abstraction is a distributed collection of items called “Resilient Distributed Dataset (RDD)”. RDDs can be created from Hadoop InputFormats’
HTTP://DOCS.SCIPY.ORGIDOCINUMPY/REFERENCE/ROUTINES.LINALG.HTML
HTTP://DOCS.SCIPY.ORGICLOCISCIPYIREFERENCE/SPARSE.HTML
or by transforming other RDDs. Numpy arrays, Python list and Scipy CSC sparse matrices are all supported. In addition: MLIB, the Spark library for machine learning, supports SparseVectors and LabeledPoint, i.e. local vectors, either dense or sparse, associated with a label/response
Code
import numpy as np
from scipy.sparse import csr_matrix
M = csr_matrix ([[4, 1, 0], [4, 0, 3], [0, 0, 1]])
from pyspark.mllib.linalg import SparseVector
from pyspark.mllib.regression import LabeledPoint
label = 0.0 point = LabeledPoint(label. SparseVector(3, [0, 2], [1.0, 3.0]))
textRDD = sc.textFile(“README.md”)
print textRDD.count() # count items in RDD

Q. Can you provide an example for Map and Reduce in Spark? (Let’s compute the Mean Square Error)
Spark is a powerful paradigm for parallel computations which are mapped into multiple servers with no need of dealing with low level operations such as scheduling, data partitioning, communication and recovery. Those low level operations were typically exposed to the programmers by previous paradigms. Now Spark solves these problems on our behalf.
A simple form of parallel computation supported by Spark is the “Map and Reduce” which has been made popular by Google.
HTTP://RESEARCH.GOOGLE.COM/ARCHIVE/MAPREDUCE.HTML

In this framework a set of keywords is mapped into a number of workers (e.g. parallel servers available for computation) and the results are then reduced (e.g. collected) by applying a “reduce” operator. The reduce operator could be very simple (for instance a sum) or sophisticated (e.g. a user defined function).
As an example of distributed computation let’s compute the Mean Square Error (MSE), the average of the squares of the difference between the estimator and what is estimated.
In the following example we suppose that valuesAndPreds is an RDD of many (vi = true labels, pi = predictions) tuples. Those are mapped into values (vi – pi)2. All intermediate results computed by parallel workers are then reduced by applying a sum operator. The final result is then divided by the total number of tuples as defined by the mathematical definition MSE = 1/n nEi=1 (vi – pi)2. Note that Spark hides all the low level details to the programmer by allowing to write a distributed code which is very close to a mathematical formulation.
Code
MSE = valuesAndPreds.rnap(lambda (v, p): (v – p)**2).reduce(lanthda x, y: x + y) / valuesAndPreds.count()
Spark can however support additional forms of parallel computation by taking inspiration from the 20 years of work on skeletons computations and, more recently, on Microsoft’s Cosmos.

Q. Can you provide examples for other computations in Spark?
The first code fragment is an example of map reduction, where we want to find the line with most words in a text. First each line is mapped into the number of words it contains. Then those numbers are reduced and the maximum is taken. Pretty simple: one single line of code stays here for something which requires hundreds of lines in other parallel paradigms such as Hadoop.
Spark supports two types of operations: transformations, which create a new RDD dataset from an existing one, and actions, which return a value to the driver program after running a computation on the dataset. All transformations in Spark are lazy because they postpone computation as much as possible until the results are really needed by the program. This allows Spark to run efficiently — for example the compiler can realize that an RDD created through map will be used in a reduce and return only the result of the reduce to the driver, rather than the larger mapped dataset. Intermediate results can be persisted and cached.
Basic transformations include (the list below is not comprehensive. Check online for a full list)

Transformation	Use
Map(func)	Returns a new distributed dataset formed by passing each element of the source through a function func.
filter(func)	Returns a new dataset formed by selecting those elements of the source on which func returns true.
flatMap(func)	Similar to map, but each input item can be mapped to 0 or more output items (so func should return a Seq rather than a single item).
sample(withReplacement, fraction, seed)	Samples a fraction of the data, with or without replacement, using a given random number generator seed.
union(otherDataset)	Returns a new dataset that contains the union of the elements in the source dataset and argument.
intersection(otherDataset)	Returns a new ROD that contains the intersection of elements in the source dataset and argument.
distinct([numTasks]))	Returns a new dataset that contains the distinct elements of the source dataset.
groupByKey([numTasks])	When called on a dataset of (K, V) pairs, a dataset of (K, Iterable) pairs returns.
reduceByKey(func, [numTasks])	When called on a dataset of (K, V) pairs, a dataset of (K, V) pairs returns, where the values for each key are aggregated using the given reduce function func, which must be of type (V,V) => V
sortByKeyflascending], [numTasks])	When called on a dataset of (K, V) pairs, where K implements Ordered, a dataset of (K, V) pairs sorted by keys in ascending or descending order returns, as specified in the Boolean ascending argument.

HTTP://CODINGPLAYGROUND.BLOGSPOT.COAT/2010/09/COSMOS-MASSIVE-COMPUTATION-IN-ICROSOFT.HTML
HTTP://SPARK.APACHE.ORGICLOCSILATESTIPROGRAMMING-
GUIDE.HTMUTTRANSFORRNATIONS

120 Data Science Interview Questions

Here are the answers to 120 Data Science Interview Questions.

• Communication
• Data Analysis
• Predictive Modeling
• Probability
• Product Metrics
• Programming
• Statistical Inference