ML noobs

Everything you need to know about TensorFlow Certification Exam

Prathamesh Parit — Mon, 31 Oct 2022 13:42:54 GMT

So recently on 30th October 2022, I passed my Tensorflow Certification Exam and in this blog, I'll share my entire journey from preparation to getting certified, starting with

Resources I used to prepare for the exam:

Deep learning Specialization (Andrew Ng) - Coursera
TensorFlow Developer Certificate in 2022 - Zero To Mastery
DeepLearning.AI TensorFlow Developer Professional Certificate

also, some internships did help me boost my skills in the optimization part.

Of course, you are not limited to only these resources. It's just that these helped me and something else might work for you.

Actually, my entire certification exam was sponsored by Google as I was a part of Google Machine Learning Bootcamp 2022 Cohort 1 and I was provided with a coupon to undertake the examination.

After you are confident enough that you are ready for the exam, you can appear the exam using this link.

But before that, here are some things you need to know about the exam:

Go through the TensorFlow Developer Certificate Candidate Handbook to clear most of the queries related to the exam.

It is advisable to sign up 2 hours before the exam as the ID verification might take up to 5 minutes to 2 hours to verify. You must have any of the following IDs for ID verification
- Passport
- Voting Id
- Driving License
- Identity Card
It is usually an automated process where they'll tell you within 5 mins whether your id is verified or not, and if not it'll take about 2 hours to verify the process by their admin manually.
During filling out your personal details even though if you have the coupon code you still have to enter your billing details and afterwards it'll just say that amount is $0 so you have nothing to pay.
After payment is done then you get an option saying to redeem the exam. You redeem it and a start option will pop up then you can start your exam.

Things you need to know to start the exam:

You need to get familiar with Pycharm IDE
You to install the "TensorFlow Certification Exam" Plugin and this link will explain the steps to install the plugin and how to start the exam

Things you need to know after you start the exam:

As shown above in the image you'll be provided with 5 categories with each category containing 1 question.

The categories may consist of 5 types:
1. Neural Networks (Classification/Regression)
2. Image Classification (Binary/ Multiclass)
3. Computer vision with CNNs
4. NLP
5. Time series, Sequences, and Prediction
Model training can be done using Pycharm and mostly the dataset for these questions are not huge and are probably compatible to train on your local device but still if your PC doesn't have enough computational power to train the model then you can train your model on the colab and download the model in "modelname.h5" format and paste it in the respected category folder. They don't check the code but only evaluate the saved model on their private test cases.
There are 5 questions in this exam with increasing difficulty from 1-5. Please note that the weight of the grade for the question is relative to its difficulty. So your Category 1 question will score significantly less than your Category 5 question.
After the model is trained and saved, you can test and submit the model using the Assistant provided on the right side of Pycharm. (Note: You can test and submit your model multiple times and there is no penalty for that so you can test it multiple times until you score 5/5)
It is a 5-hour long exam and you need to score above 90% to pass the exam. As the weightage differs according to the difficulty it is advisable to score 5/5 on each question.

Types of questions I received:

Note: To preserve the confidentiality and excitement of the exam these questions won't be exactly same but similar to this

Getting Started Question

A basic regression question dataset will be given and we have to train a neural network to match the x to the y.
Basic Datasets Question

Create and train a classifier for the Tensorflow Image dataset. Note that the test will expect it to classify n classes and that the input shape should be the native size of the TF dataset which is NxN monochrome. Do not resize the data. Your input layer should accept NxN as the input shape only. If you amend this, the tests will fail.
Computer vision with CNNs

Create and train a classifier for CV Image Dataset using the provided data. The test will use images that are NxN with N bytes color depth so be sure to design your neural network accordingly
NLP QUESTION

Build and train a classifier for the NLP Dataset. It will be tested against a number of sentences that the network hasn't previously seen and you will be scored on whether prediction was correctly detected in those sentences.
TIME SERIES QUESTION

Build and train a neural network to predict the time-indexed variable of the dataset_name Using a window of past N observations of N feature, train the model to predict the next N observations of that feature.

Things you need to know after you End the exam:

You will have an End test option on the top right side of pycharm.
After ending, immediately a mail will be sent saying whether you are pass or not(No marks are shown)
If you pass, another Tensorflow certification mail will be sent along with your certificate and badge that you have passed the Tensorflow Developer Exam.
Also google forms will be sent to add your name to the Tensorflow Network where recruiters will find out a Tensorflow developer in a particular region.

My personal experience and tips to qualify for the exam

To be honest I faced multiple difficulties during the exam I was completely unaware about the fact that how the exam is going to be for example details about plugins and Assistant. So to help people to overcome these problems this blog is going to help you a lot to get an overview of exam.

Some of those difficulties were that my pycharm was not compatible with the TF Developer Exam Plugin so I had to reinstall the entire setup.

Also, the virtual environment that I was provided was causing multiple module errors which it wasted my 30mins and I used another interpreter for the exam. (Note: Though you only need a saved model to pass the test and not pycharm code so even if you couldn't do it, you can download the model from colab)

For me, the exam was completed in the first 2-hours itself but it took me 3-hours to optimize the code as I scored pretty terribly like 2/5 and 3/5 in the first tries, and after few hours of optimization I went up to 5/5

My scores were:

Question 1: 5/5
Question 2: 5/5
Question 3: 5/5
Question 4: 4/5
Question 5: 5/5

And at last, when I passed the exam it felt like I was at the top of the world and it's just the beginning 💯. Every bit of hard work and time I invested with consistency was totally worth the satisfaction I felt at last☺️.

Tips:

I would rate the exam somewhere between medium and hard.
It is not about building the entire model as most of the time the entire model will be already present. You just have to write the code between sequential, compiling, and fitting the model.
Focus more on the optimization of code as it is only the part where you score.

Supervised vs Unsupervised (Classification)

Prathamesh Parit — Sun, 06 Mar 2022 16:39:26 GMT

2. Classification Model:

Trained on the training dataset and based on that training, it categorizes the data into different classes.

Example: The best example to understand the Classification problem is Email Spam Detection. The model is trained on the basis of millions of emails on different parameters, and whenever it receives a new email, it identifies whether the email is spam or not. If the email is spam, then it is moved to the Spam folder.

Here OUTPUT is always DISCRETE.

Types:

1.1 Logistic Regression:

The output values can only be between 0 and 1

DIFFERENCE BETWEEN LOGISTIC AND LINEAR REGRESSION IS:

Values of Linear Reg can exceed 0 and 1 range, while values of Logistic Reg lies within 0 and 1 range

1.2 Naive Bayes Classifier:

It is a probabilistic classifier, which means it predicts on the basis of the probability of an object.

Naïve: It is called Naïve because it assumes that the occurrence of a certain feature is independent of the occurrence of other features. Such as if the fruit is identified on the bases of color, shape, and taste, then red, spherical, and sweet fruit is recognized as an apple. Hence each feature individually contributes to identify that it is an apple without depending on each other.

Bayes: It is called Bayes because it depends on the principle of Bayes' Theorem.

Where,

P(A|B) is Posterior probability: Probability of hypothesis A on the observed event B.

P(B|A) is Likelihood probability: Probability of the evidence given that the probability of a hypothesis is true.

P(A) is Prior Probability: Probability of hypothesis before observing the evidence.

P(B) is Marginal Probability: Probability of Evidence.

1.3 K-Nearest Neighbor:

Let’s say we want to classify the given point into one of the three groups.

In order to find the k nearest neighbors of the given point, we need to calculate the distance between the given point to the other points.

Then, we need to sort the nearest neighbors of the given point by the distances in increasing order.

For the classification problem, the point is classified by a vote of its neighbors, then the point is assigned to the class most common among its k nearest neighbors.
# knn from sklearn
from sklearn import neighbors, datasets
# import some data to play with
iris = datasets.load_iris()
# we only take the first two features for demonstration
X = iris.data[:, :2]
y = iris.target
clf = neighbors.KNeighborsClassifier(n_neighbors=15)
clf.fit(X,y)
The left plot shows classification decision boundary with k = 15, and the right plot is for k = 3.

1.4 Support Vector Machine:

Dots = Features

SVMs perform the classification test by drawing a hyperplane that is a line in 2d or 3d

In such a way that all points of one category are on one side of the hyperplane and all points of the other category are on the other side

There could be multiple such hyperplanes

Distance between 2 categories is called the margin and the points that fall exactly on the margin are called the supporting vector

What is Machine Learning?

Prathamesh Parit — Fri, 11 Feb 2022 14:37:02 GMT

Machine Learning:

Machine learning is turning things (data) into numbers and finding patterns in those numbers. We can use ML for literally anything as long as you can convert it into numbers and program it to find patterns. Literally, it could be anything any input or output from the universe.

Again Machine Learning is divided into two parts:

Basically, machine learning teaches a computer how to perform a task without explicitly programming it to perform the said task instead of feeding data into an algorithm to gradually improve outcomes with an experience similar to how organic life learns.

How does Machine learning work in the backend?

Step 1: Acquire and Clean up data

Lots and lots of data the better

Data needs to have some kind of signal to be valuable to the algorithm for making predictions.

To transform raw data into features that better represent the underlying problem

Step 2: Separate data into a training and testing set

the training data is fed into an algorithm to build a model

the testing data is used to test the accuracy or error of the model to check how it performs on unseen data

Step 3: Choose an Algorithm

Basically, you can't just train every data with the same model. Every problem so it needs different models

Like for predicting numbers we use REGRESSION. (Eg: Predicting Stock Prices)

Predicting between two things we use CLASSIFICATION (Eg: Predicting the between cat and dog)

Supporting Frameworks(Softwares):

numpy
pandas
matplotlib
scikit-learn
tensorflow

Conclusion:

machine-learning is a model which is just a file that takes some input data in the same shape that it was trained on then spits out a prediction that tries to minimize the error that it was optimized for it can then be embedded on an actual device or deployed to the cloud to build a real world product

Supervised vs Unsupervised (Regression)

Prathamesh Parit — Thu, 10 Feb 2022 18:30:00 GMT

Machine Learning Models:

Supervised Learning:

Involves a series of function that map's an to an output based on a series of example input-output pairs.

Eg: We have 2 variables Age(Input) and Shoe size(Output) we can predict shoe size of people acc to age using supervised leaning model

SUPERVISED LEARNING there are two sub categories one is REGRESSION and other is CLASSIFICATION

1. Regression Model:

Find relationship between a dependent Variable and Independent Variable

Here OUTPUT is always CONTINOUS.

Types:

1.1 Linear Regression:

Finding a curve/ line for best fit

Right answer is on the red line and blue dots are the predictions made by model

1.2 Decision Tree:

To create a training model that can use to predict the class or value of the target variable by learning simple decision rules inferred from prior data(training data).

1.3 Random Forests:

Random Forest are so called because each tree in the forest is built by randomly selecting a sample of the data

Relies on majority wins model it reduces the risk of error from individual tree.

For example, the prediction for tress 1 and 2 is apple. Another decision tree (n) has predicted banana as the outcome. The random forest classifier collects the majority voting to provide the final prediction. The majority of the decision trees have chosen apple as their prediction

1.4 Neural Network:

Using algorithms, they can recognize hidden patterns and correlations in raw data, cluster and classify it, and – over time – continuously learn and improve.

Is a multi layered model inspired by human minds like neurons in our brain the circle represents a node