Machine Learning & Neural Networks- 2026 Solved Problems

1. In 2026, which architecture remains the backbone of most Large Language Models (LLMs) due to its "Self-Attention" mechanism? A) Recurrent Neural Networks (RNN) B) Transformer Architecture C) Convolutional Neural Networks (CNN) D) Support Vector Machines (SVM) Correct Answer: B) Transformer Architecture 2. What is "Overfitting" in a Machine Learning model? A) When the model is too small for the data B) When the model performs well on training data but poorly on unseen test data C) When the training process takes too much time D) When the model only uses 10% of the available RAM Correct Answer: B) When the model performs well on training data but poorly on unseen test data 3. Which process involves training a model on a large dataset and then fine-tuning it for a specific, smaller task? A) Linear Regression B) Transfer Learning C) Data Scraping D) Feature Scaling Correct Answer: B) Transfer Learning 4. What is the primary function of an "Activation Function" (like ReLU or Sigmoid) in a Neural Network? A) To delete unnecessary data B) To introduce non-linearity into the model C) To increase the number of layers D) To speed up the internet connection Correct Answer: B) To introduce non-linearity into the model 5. In 2027, "Reinforcement Learning from Human Feedback" (RLHF) is primarily used to: A) Make models run faster B) Align AI behavior with human values and preferences C) Reduce the cost of GPUs D) Automatically write code for websites Correct Answer: B) Align AI behavior with human values and preferences 6. Which type of learning uses labeled data (input-output pairs) for training? A) Unsupervised Learning B) Supervised Learning C) Self-supervised Learning D) Zero-shot Learning Correct Answer: B) Supervised Learning 7. "Gradient Descent" is an optimization algorithm used to: A) Increase the error rate B) Minimize the Loss Function by adjusting weights C) Sort data in alphabetical order D) Create new data points Correct Answer: B) Minimize the Loss Function by adjusting weights 8. What is a "Neural Network Layer" that is neither the input nor the output layer called? A) Shadow Layer B) Hidden Layer C) Ghost Layer D) Middle-man Layer Correct Answer: B) Hidden Layer 9. In computer vision, which type of network is most effective for image recognition? A) Long Short-Term Memory (LSTM) B) Convolutional Neural Networks (CNN) C) Linear Discriminant Analysis (LDA) D) Decision Trees Correct Answer: B) Convolutional Neural Networks (CNN) 10. What is "Backpropagation"? A) Moving data from the cloud to a local drive B) The method used to calculate gradients and update weights in a neural network C) A way to delete old models D) The process of printing a neural network diagram Correct Answer: B) The method used to calculate gradients and update weights in a neural network 11. Which term describes the gap between the predicted value and the actual value? A) Accuracy B) Loss (or Error) C) Precision D) Recall Correct Answer: B) Loss (or Error) 12. "Unsupervised Learning" is commonly used for which task? A) Predicting house prices B) Clustering (grouping similar data points without labels) C) Email spam detection D) Image classification Correct Answer: B) Clustering (grouping similar data points without labels) 13. What does "Epoch" mean in Machine Learning training? A) One full pass of the entire training dataset through the neural network B) The date the model was created C) A period of 100 years D) The time it takes to save a file Correct Answer: A) One full pass of the entire training dataset through the neural network 14. In 2026, "Quantization" is a popular technique to: A) Make the model smarter B) Reduce the size and memory usage of a model for use on mobile devices C) Increase the number of parameters D) Create high-resolution images Correct Answer: B) Reduce the size and memory usage of a model for use on mobile devices 15. "Hyperparameters" are: A) The data used for training B) Settings configured before the training process (e.g., Learning Rate, Batch Size) C) The final results of the model D) The speed of the processor Correct Answer: B) Settings configured before the training process (e.g., Learning Rate, Batch Size)