Level3_viva20 MLP Level 1 Aug 7, 2026 Copy link 1 Explain notebook 2 Explain how PCA works 3 Did you do hyper parameter tuning
Viva18 MLP Level 1 Aug 7, 2026 Copy link 1 Show your ID card. 2 Explain your problem statement. 3 Explain your notebook/project. 4 How did you approach the problem? 5 Explain your EDA. 6 What did you show in the graphs? 7 Explain the skewness of the target variable. 8 Explain your preprocessing steps. 9 Explain your feature engineering. 10 Explain your model training. 11 Which model did you use? 12 How can you improve model performance? 13 Explain your hyperparameter tuning. 14 Show your hyperparameter tuning implementation. 15 Explain your parameter grid. 16 What do these hyperparameters mean? colsample_bytree learning_rate max_depth n_estimators subsample 17 What happens if the learning rate is increased? 18 What happens if the number of estimators is increased? 19 What happens if the maximum depth is changed? 20 Show your Pipeline. 21 What is TSCV (Time Series Cross Validation)? 22 Why did you use TSCV? 23 What does the cv parameter mean in TSCV? 24 Why can your dataset be considered time-series data? 25 Why do we perform Cross Validation? 26 Why did you use the rolling mean feature? 27 What data/features did you not use, and why? 28 What is the shape of the training dataset? 29 What is the shape of the test dataset? 30 Print the first five rows of the training dataset. 31 Show the basic statistics of the training dataset using describe(). 32 Count the null values in the training dataset.
Level3_20 MLP Level 1 Aug 7, 2026 Copy link 1 Show your ID card. 2 Introduction. 3 Explain your notebook/project. 4 How many models did you train? 5 On how many models did you perform Hyperparameter Tuning? 6 Why did you perform Hyperparameter Tuning only on those models? 7 Which model took the longest to train? Why? 8 Explain Gradient Descent. 9 Explain Gradient Descent on a whiteboard. 10 Why did you use Accuracy instead of F1-Score or ROC-AUC? 11 Draw and explain the Confusion Matrix. 12 Write and explain the formulas for Precision, Recall, F1-Score, and Accuracy. 13 How did you perform feature engineering on date/time features? 14 Which hyperparameters did you tune? 15 What time series model would you use? 16 What more could you have done to improve the project? 17 How did you handle class imbalance? 18 What does the C parameter in Logistic Regression represent? 19 Why did you use Median instead of Mean (or vice versa) for imputation? 20 Which imputation technique would you use for a right-skewed feature? 21 Which model did you use for the final submission? 22 Show your Kaggle leaderboard score and rank. 23 Show your Kaggle submissions. 24 Questions based on your notebook implementation. 25 Suggestions on improving the notebook/project. 26 Asked if you have any questions for the proctor.
level3_30 MLP Level 1 Aug 7, 2026 Copy link 1 Introduction. 2 Explain your notebook/project. 3 How did you handle missing (NaN) values? 4 Which imputer should be used when a feature contains outliers? 5 How does Logistic Regression work? 6 What is the loss function of Logistic Regression? 7 How does Linear Regression work? 8 What is the loss function of Linear Regression? 9 How does Random Forest work? 10 How does XGBoost work? 11 Difference between Bagging and Boosting. 12 How can you handle an imbalanced dataset? 13 What is the Pearson Correlation formula? 14 Do all trees in a Random Forest receive all features? 15 What are the types of feature reduction techniques? 16 How does PCA work? 17 Is PCA linear or non-linear? 18 What are Eigenvectors? 19 How do you calculate PCA manually? 20 How do you calculate Eigenvectors manually? 21 Why does L1 Regularization eliminate features? 22 Why does L2 Regularization help prevent overfitting? 23 In a poisonous apple detection problem, would you prioritize Precision or Recall? Why?
level1-8 MLP Level 1 Aug 7, 2026 Copy link 1 Explain your EDA. 2 Explain your EDA code. 3 Suggestions to improve the notebook/code. 4 No additional theory or coding questions were asked.
level1_viva2 MLP Level 1 Aug 7, 2026 Copy link 1 Show your ID card. 2 Introduction. 3 Tell me about yourself. 4 Tell me about yourself. 5 Explain the problem statement. 6 Explain your notebook/project. 7 Explain your preprocessing. 8 Explain your EDA. 9 Explain your feature engineering. 10 Explain your hyperparameter tuning. 11 Explain your model comparison. 12 What insights did you gain from your analysis? 13 What difficulties did you face during the project? 14 How did you overcome those difficulties? 15 What did you learn from the project? 16 Show an overview of your Kaggle submissions. 17 What models did you use? 18 Why did you use those models? 19 Why did you use only tree/leaf-based models? 20 Why do we perform hyperparameter tuning? 21 What are hyperparameters? 22 Difference between parameters and hyperparameters. 23 How is hyperparameter tuning done? 24 Ways to perform hyperparameter tuning. 25 Difference between Pipeline and ColumnTransformer. 26 Difference between Pipeline and Transformer. 27 Why did you use StandardScaler? 28 Which scaling method is better for handling outliers? 29 Why do machine learning models use numerical features? 30 What is PCA? 31 Difference between XGBoost and LightGBM. 32 What is overfitting? 33 How do you reduce overfitting? 34 What is the Bias-Variance Tradeoff? 35 Difference between Bagging and Boosting. 36 Which sklearn API contains OneHotEncoder? 37 Which sklearn API contains StandardScaler? 38 Coding: Implement the preprocessing Pipeline used in your notebook. 39 Coding: Write a Decision Tree classifier snippet. 40 Coding: Load the California Housing dataset. 41 Coding: Perform a train-test split on the California Housing dataset. 42 Coding: Load the Iris dataset. 43 Coding: Perform a train-test split on the Iris dataset. 44 Coding: Train a Decision Tree classifier on the Iris dataset. 45 Coding: Print the Accuracy Score. 46 Coding: Use the describe() function on the California Housing dataset.
mlp_level1_viva9 MLP Level 1 Aug 7, 2026 Copy link 1 Show your ID card. 2 Explain your notebook/project. 3 Explain your EDA. 4 Explain your univariate analysis. 5 Explain your bivariate analysis. 6 Explain the graphs you created and the insights from them. 7 How did you handle missing values? 8 What is SimpleImputer? 9 How does SimpleImputer work? 10 Explain your preprocessing pipeline. 11 Have you used Pipeline? 12 Why did you use (or not use) Pipeline? 13 Explain your feature engineering. 14 Did you create any new features? 15 What encoding technique did you use? 16 Why did you choose that encoder? 17 Replace OrdinalEncoder with OneHotEncoder. 18 Difference between Label Encoding and OneHot Encoding. 19 What scaler did you use? 20 Why did you choose MinMaxScaler over StandardScaler? 21 Explain StandardScaler. 22 Difference between StandardScaler and MinMaxScaler. 23 What models did you try? 24 Which model performed the best? 25 Why didn't you explore linear models? 26 Explain your model comparison. 27 Explain the hyperparameters you tuned. 28 What is the learning rate? 29 Is a higher or lower learning rate better? Why? 30 What is the Bias-Variance Tradeoff? 31 What is Boosting? 32 What is Gradient Boosting? 33 What algorithm does Boosting use? 34 What is Bagging? 35 Difference between Bagging and Boosting. 36 Difference between XGBoost and LightGBM. 37 What is R² Score? 38 Write the formula for R² Score. 39 From which library do you import r2_score? 40 From which library do you import OneHotEncoder? 41 From which library do you import XGBoost? 42 Why is Logistic Regression called "Regression" in a classification task? 43 What is F1-Score? 44 Write the formulas for Accuracy, Precision, and F1-Score. 45 What are Ensemble methods? 46 What resources did you use while building the project? 47 Coding: Create a dataframe containing only categorical (object) columns. 48 Coding: Apply OneHotEncoder to categorical columns with ≤10 unique values and find the new number of columns. 49 Coding: Split the training data into numerical and categorical dataframes. 50 Coding: Build separate preprocessing pipelines for numerical and categorical features using ColumnTransformer. 51 Coding: Perform a train-test split. 52 Coding: Print all numerical and non-numerical columns. 53 Coding: Print Recall Score instead of Accuracy Score. 54 Coding: Implement RandomForestRegressor (import, initialize, fit, predict). 55 Coding: Build a simple preprocessing pipeline. 56 Coding: Write a custom preprocessing pipeline. 57 Coding: Apply Label Encoding to ["green", "blue", "white", "blue", "green"]. 58 Coding: Filter rows where RegionCode = "Florida". 59 Coding: Filter rows where RegionCode = "Florida" and TargetValue > 50000. 60 Coding: Create a subset of the data for RegionCode = "Florida". 61 Coding: Drop two specified columns from the training dataset.
level1_viva8_MLP_Viva MLP Level 1 Aug 7, 2026 Copy link 1 Introduction 2 Start explaining your notebook from the EDA section. 3 Explain every line of your code in detail. 4 No additional theory questions were asked.