Neural Networks
This is the basic building block of a neural network, inspired by biological neurons.
What is a perceptron (or artificial neuron)?
This technique transforms features to have a mean of 0 and standard deviation of 1, helping models converge faster.
What is Normalization (or Standardization)?
This type of learning uses labeled data with input-output pairs.
What is Supervised Learning?
The process by which a neural network adjusts its weights based on the error of its predictions.
What is training (or learning)?
This supervised learning algorithm predicts continuous values by fitting a straight line to the data.
What is linear regression?
This activation function is commonly used in hidden layers and helps prevent the vanishing gradient problem.
What is ReLU (Rectified Linear Unit)?
This method converts categorical variables into binary vectors where only one element is 1 and the rest are 0.
What is One-Hot Encoding?
This metric measures the proportion of correct predictions made by a model, calculated as correct predictions divided by total predictions.
What is Accuracy?
This measure of a model's ability to generalize is calculated by evaluating performance on data the model has never seen during training.
What is validation/test accuracy?
This classification algorithm uses a sigmoid function to predict probabilities between 0 and 1.
What is logistic regression?
This technique randomly drops units during training to prevent overfitting and improve generalization.
What is Dropout?
This technique creates synthetic training examples by slightly modifying existing data, commonly used in computer vision with rotations and flips.
What is Data Augmentation?
This process divides a dataset into training, validation, and test sets to evaluate model performance.
What is Train-Test Split (or Data Splitting)?
This term refers to one complete pass through the entire training dataset during the training process.
What is an epoch?
This tree-based model makes predictions by splitting data based on feature values, creating a flowchart-like structure of decisions.
What is a decision tree?
This algorithm propagates the error backwards through the network, adjusting weights to minimize loss.
What is Backpropagation?
This technique uses statistical methods to replace missing values in a dataset, often using mean, median, or mode.
What is Imputation?
This phenomenon occurs when a model learns the training data too well, including its noise, leading to poor performance on unseen data.
What is Overfitting?
This hyperparameter controls how much the model's weights are adjusted with respect to the loss gradient during training.
What is learning rate?
This ensemble method combines multiple decision trees and uses voting to make final predictions, reducing overfitting.
What is a random forest?
These specialized neural networks use convolutional layers with filters to automatically learn hierarchical features from spatial data like images.
What are Convolutional Neural Networks (CNNs)?
This advanced technique generates new synthetic training samples by interpolating between existing data points, often used to address class imbalance.
What is SMOTE (Synthetic Minority Over-sampling Technique)?
This evaluation metric combines precision and recall into a single score, particularly useful when dealing with imbalanced datasets.
What is F1-score?
This attention mechanism, introduced in the Transformer architecture, allows the model to weigh the importance of different parts of the input sequence.
What is self-attention (or attention mechanism)?
This DeepMind AI system defeated world champion Lee Sedol at the ancient board game Go in 2016, using deep reinforcement learning and Monte Carlo tree search.
What is AlphaGo?