🔵 Aggregating Data Science : My Newsletter📩Manralai : AI Career Master

This repository serves as a comprehensive log of my contributions to the field of data, as well as a testament to my unwavering commitment to making complex concepts accessible to others through clear and concise explanations

---

This is Mukesh Manral

🔴Statistics

Framework to Select Statistical Test

Statistics	Understant these Basics(will update)
📙 Descriptive Statistics	* Measure of Frequency and Central Tendency * Measure of Dispersion * Probability Distribution * Gaussian Normal Distribution * Skewness and Kurtosis * Continuous and Discrete Functions * Goodness of Fit * Normality Test
📙 Inferential Statistics	* t-Test & it's Types * z-Test * Hypothesis Testing * Typr-I & Type-II Error * One way ANOVA * Two way ANOVA * Chi-Square Test
Statistical Test's for Data Science in Python	Link

---

🔴 0 to Python

Pyton	Guide + Video Link
📙 Python - Guide & Playlist	YT Playlist + Guide
📙 Python - Patterns Practice	YT Private Playlist
📙 Python - Functions & OOP's	YT Private Playlist
📙 Learn Python the Applied Way	YT Playlist + Notebook
📙 Python Project Outline	Private Code

🔴Python to Expert

Pyton	Guide + Video Link
📙 Algorithms - LeetCode Crux	Private Link
📙 Python - Data Structure and Algorithm Foundation	Private Link
📙 Python - Data Structure & TimeComplexities Crux	Private Link
📙 Asymptotic Notations [Big-O,Omega,Theta] Crux	Private Link
📙 Python Data Structure - Array,Strings	Private
📙 Python Coding Interview Questions	Private
📙 Python Quick Revision	Private
📙 Python Interview Questions	Private
📙 Python-RevisionQuestions	Private

---

🔴Data Analysis

Data Analysis	Code + Blog Link
📙 Mapping Business Problem into Data Science Problem	Private Link
📙 Answer to --> How to Perform EDA	Private Link
📙 Evaluation Metrics (Supervised Learning)	Private Link

🔴Machine Learning

Imbalanced Binary Classification Metrices Guide

Machine Learning	Code	Blog - Video Link
📙 Machine Learning Life Cycle	Private Link
📙 Simple to understand ML framework	Link
📙 Problem Solving[Mapping Business Problem to Data Science Problem]	Private Link
📙 First Classificaton + Regression Models - How to?? When to?? of Modeling	Private Link
📙 Logistic Regression Derivation(Probabilistic,Geometric,Loss Interpretations)	Private Link
📙 Linear Regression derivation	Private Link
📙 Linear Regression Hypothesis + interview questions	Link
📙 Linear Regression implementing Cost Function	Link
📙 Multiple Linear Regression (Basic)	Link
📙 Polynomial Regression (Basic)	Link
📙 Advance Linear Regression + GridSearchCV + HyperParameterTuning(Basic)	Link
📙 K-NN Deriving	Private Link
📙 k-NN Algorithm-1 (Basic)	Link
📙 k-NN Algorithm-2 (Basic)	Link
📙 Propose Saving Options for for India based Poultry Farm having spending of appx 22 billion	Link+KPI's
📙 Mastering Dimensionality Reduction Techniques	Link
📙 PCA application you must know things	Link	PCA - Second Principal Component have to be orthogonal to first one? See Why!
📙 PCA application step by step - 2	Link	PCA-1, PCA-2
📙 Vehicle Loan Default Prediction [EDA + Model]	Link
📙 Decision Tree Derivation	Private Link
📙 Decision Tree end2end + HyperparameterTuning	Link
📙 Decision Tree Questions	Private Link
📙 Bagging(ensemble_concept[RF-ExTree])	Link
📙 Employees-earnings (Model comparison, EDA many more)	Link
📙 SVM	Link

🔴Feature Engineering

Feature Engineering	Code + Blog Link
📙 Creating New Features	Link
📙 Imputation (categorical,numerical) + Handling Outliers (standard deviation,percentile) + Scaling (normalization, standardization) + Binning + Encoding Categorical Variable + Featur Encoding (integer,onehot))	Link
📙 Variable Transformation (Square Root, Reciprocal, Logarithmic, Exponential\Power Transformation	Link
📙 DateTime Engineering	Link

---

🔴Deep Learning

State of the Art Object Detection Timeline

Deep Learning	Code + Blog Link
📙 kNN to Parameterised Learning	Link
📙 Deep Learning & Perceptron Trick	Private Link
📙 Implementing Neural Network using NumPy from Scratch	Private Link
📙 Keras Functional Models	Link
📙 Hyper Parameter Tuning + Activation Functions + Selecting Right Activation Function + Optimizer + Loss, Error, Cost Functions + Project(Keras): Loan Prediction + Project: Image Classification(Emergency Vs Non-Emergency Vehicle) using NN	Private Link
📙 Gradient Descent from Scratch	Link
📙 SGD with Momentum from Scratch	Link
📙 Weight Initialisation Techniques	Link
📙 Activation Functions	Link
📙 Loss Functions	Link
📙 Optimizers & Fast Optimizers	Link
📙 Regularization : Avoiding Overfitting	Update Private Link
📙 Improving Deep Neural Networks: Hyperparameter Tuning,Regularization & Optimization	Link
📙 Applying KerasTuner & Dropout	Link
📙 More Implementation	Link
📙 Neural Network + ANN-Application-3 projects	Link
📙 Loan Prediction Using Neural Network in Keras	Link
📙 TensorFlow2.x & CNN	Link
📙 CNN - ConvNet	Link
📙 CNN on Cifar10	Link
📙 Image Captioning using Keras	Link
📙 TransferLearning AlexNet-(2012) - Classification	Link
📙 TransferLearning VGGNet-(2014) - Classification	Link
📙 TransferLearning ResNet-(2015) - Classification	Link
📙 Object Detection Basics	Link
📙 Deep Into Object Detection	Private Link
📙 Regional CNNs differeces (R-CNN, Fast R-CNN, Faster R-CNN)	Link
📙 R-CNN - Object Detection	Link
📙 Fast R-CNN - Object Detection	Update Link
📙 Faster R-CNN - Object Detection	Link
📙 SSD - Single Shot Detector
📙 YOLO - Single Shot Detector
📙 SceneClassificatio(On-places365_small)	Link

---

---

Scene Classification Algos : Mostly used ones

• ResNet is used by Google for scene classification in its image search engine
• EfficientNet is used by Facebook for scene classification in its photo tagging system
• MobileNet is used by Snapchat for scene classification in its augmented reality filters
• InceptionNet is used by Microsoft for scene classification in its Azure Cognitive Services

Algorithm	Strengths	Weaknesses	Best suited for
ResNet	Accurate, efficient, and robust to noise.	Can be computationally expensive to train.	Large datasets and complex scene classification tasks.
VGGNet	Very accurate, but computationally expensive.	Can be overfitting to small datasets.	Medium-sized datasets and simple scene classification tasks.
GoogleNet	Accurate and efficient, but not as robust to noise as ResNet.	Can be overfitting to small datasets.	Medium-sized datasets and simple scene classification tasks.
DenseNet	Accurate and efficient, but can be overfitting to small datasets.	Medium-sized datasets and simple scene classification tasks.
InceptionNet	Very accurate, but computationally expensive.	Can be overfitting to small datasets.	Large datasets and complex scene classification tasks.
Xception	Accurate and efficient, but not as robust to noise as ResNet.	Medium-sized datasets and simple scene classification tasks.
MobileNet	Very efficient, but not as accurate as some of the other algorithms.	Mobile devices and other resource-constrained environments.
EfficientNet	Accurate and efficient, but can be overfitting to small datasets.	Medium-sized datasets and simple scene classification tasks.

While choosing an algorithm for Scene Classification, consider following factors:

• Size and complexity of the dataset: More complex algorithms, such as ResNet and InceptionNet, typically require larger datasets to train effectively
• Computational resources available: Some algorithms, such as VGGNet and InceptionNet, can be computationally expensive to train and run
• Specific application: For example, if algorithm will be deployed on a mobile device, then a lightweight and efficient algorithm, such as MobileNet or EfficientNet, may be best choice

---

---

Object Detection Algos : Mostly used ones

• YOLO is used by Google for object detection in its self-driving car project
• EfficientDet is used by Facebook for object detection in its video surveillance system
• Faster R-CNN is used by Amazon for object detection in its warehouse automation system
• Cascade R-CNN is used by Microsoft for object detection in its Azure Cognitive Services
• Mask R-CNN is used by Tesla for object detection in its self-driving car project
• SwinT and ViT are still under development

Algorithm	Strengths	Weaknesses	Best suited for
YOLO	Fast, accurate, can be deployed on mobile devices	May not be as accurate as other algorithms for complex tasks	Real-time object detection, object detection on mobile devices
SSD	Fast, accurate, efficient use of computational resources	May not be as accurate as other algorithms for complex tasks	Real-time object detection, object detection on resource-constrained devices
EfficientDet	Fast, accurate, efficient use of computational resources	May not be as accurate as other algorithms for complex tasks	Real-time object detection, object detection on resource-constrained devices
Faster R-CNN	Accurate, can be scaled to different sizes and speeds	Can be slower than other algorithms	Object detection on high-performance devices
Cascade R-CNN	Very accurate, can be scaled to different sizes and speeds	Can be slower than other algorithms	Object detection on high-performance devices for complex tasks
Mask R-CNN	Accurate, can segment objects in images	Can be slower than other algorithms	Object detection and segmentation on high-performance devices

---

---

🔴Computer Vision + OpenCV CNN CheatSheet

Computer Vision	Code + Blog Link
📙 Image Transformation : (Pixel Manipulation, Getting & Setting Pixels, Image Translation, Rotation, Interpolation Methods and comarison, Fliping, Croping, Arithmetic, Bitwise, Masking, Channels Splitting & Merging )	Link
📙 Image Processing Techniques : (Image Enhancement,Image Restoration,Image Segmentation,Object Detection,Image Compression,Image Manipulation,Image Generation,Image-to-Image Translation)	Link
📙 Image Processing Operations : (Morphological Operations, Smoothing & Blurring, ColorSpaces, Thresholding, Adaptive Thresholding, Kernels, Image gradient-Shobel & Scharr Kernels, Canny Edge Detection, Automatic Edge Detection)	Link
📙 Morphological Operations(OpenCv)	Link
📙 Phases of Image Processing, using flow_from_directory	Link
📙 Image Histograms	Link
📙 OpenCV-Building Glasses Filter	Link
📙 Neural Networks & Parameterized Learning	Link

---

🔴NLP : RNN Cheatsheet

RNN architectures depending on task

• Recurrent Neural Networks (RNN)
  • Vanilla RNN
  • Long Short-term Memory (LSTM) -- NOTE: Understant this and all otehr are just variation
  • Peephole LSTM
  • Gated Recurrent Unit (GRU)
  • Bidirectional LSTM (Bi-LSTM)
  • Stacked LSTM
• Attention
  • Seq2seq with Attention
  • Self-attention
  • Multi-head Attention
• Transformer
  • Step 1. Adding Positional Encoding to Word Embeddings
  • Step 2. Encoder: Multi-head Attention and Feed Forward
  • Step 3. Decoder: (Masked) Multi-head Attention and Feed Forward
  • Step 4. Classifier

NLP	Code + Blog Link
(NLP Basics-1) : Tokenization,Lemmatization,Stemming,BOW,TF-IDF, N-Gram,POS,Word Embedding(Word2Vec),Gensim,CBOW,SkipGram,FastText	Link
(NLP Basics-2 Implementation) : (TextDataPreprocessing,ToLower, RemoveHTML,RemovePunctuations,HandlingChat, IncorrectTextHandling,StopWordRemoval,HandlingEmoji, Tokenization,RegularExpression,Stemmer,Lemmatization,TextRepresentation,BOW,N-gram,TF-IDF)	Link
Textual Evaluation Metrics(CER & WER),(Perplexity),(BLEU),(GLUE)	Private Link
RNN Baics(short term memory), Types, Problems and Solutions	Link
RNN Maths+Implementation	Link
RNN Architecture + Embedding	Link
LSTM	Link
RNN_LSTM_GRU Implementation + HyperBand	Link
GAN's from Scratch	Link
Transformers	Private Link

---

🔴Recommender System

Recommender System	Code + Blog Link
📙 Recommendation System Basic-(surprise_library) + All Why,What,When not, When to covered + Maths Covered +Project: Movie Recommendation(Non Personalised Recommender Systems using Count of Ratings (4 & above) + Weak Personalisation using Gender Information) + Personalised Recommender Systems + Collaborative Filtering & Types + Application of User-Based Collaborative Filtering from Scratch + Application of User Based Collaborative Filtering using [ Surprise Library ] + [ Scalability Challange ] for User Based Collaborative Filtering + Item-Based Collaborative Filtering + Implementation Item-Based Collaborative Filtering + Challenges for Neightbourhood Based Methods + Rating Prediction using [ Matrix Factorization with SVD ] + in progress	Link
📙 Why Business (H&M) cares about Product Recommendation??	Link
📙 End to End Recommender System	Link

---

🔴Spark Ecosystem

Spark Ecosystem	Code + Blog Link
📙 Spark Ecosystem(Big Data + Apache Hadoop[HDFS,YARN,MapReduce] + Spark Ecosystem[Architecture,Cluster Manager,Running application - Locally and on YARN,SparkContext vs SparkSession,spark Dataframe+Execution]+spark sql api+spark ml)	Link

---

🔴AWS

AWS	Blog Link	Video Link
📙 Cloud Computing(AWS)	Blog	YT Play List
📙 AWS Sagemaker	Blog	YT Play List

---

🔴Deployment

Deployment	Code + Blog Link	Video Link
📙 Deploy ML + Dl models using Colab & AWS	Link	---

---

🔴MLOps

• Model in Production
• Testing and Monitoring Deployments

Connect With me

🎈AI NewsLatter
🎈LinkedIn

---

© 2023 Mukesh'Manral Powered by Manralai.in