ai study notes

Activation Function

Activation Function

A mathematical function that introduces non-linearity into a neural network, determining the output of a neuron.

Backpropagation

An algorithm for calculating gradients of the loss function with respect to network parameters, enabling optimization.

Convolutional Neural Network (CNN)

A neural network architecture designed for processing grid-like data, commonly used in image recognition.

Dense Layer

A fully connected layer in a neural network where each neuron connects to every neuron in the previous layer.

Dropout

A regularization technique where randomly selected neurons are ignored during training, preventing overfitting.

Embedding Layer

A layer that maps categorical variables to low-dimensional continuous vectors, capturing relationships and similarities.

Epoch

One complete pass through the entire training dataset during model training.

GPU (Graphics Processing Unit)

A specialized processor designed for parallel computations, accelerating deep learning tasks.

Loss Function

A function that measures the error between the model's predictions and the actual target values.

Overfitting

When a model learns the training data too well, failing to generalize to unseen data.

Parameter Sharing

Using the same weights and biases for different parts of the input data, common in CNNs.

Recurrent Neural Network (RNN)

A neural network designed for processing sequential data, maintaining a hidden state to capture temporal dependencies.

Regularization

Techniques to prevent overfitting, such as weight decay or dropout.

Softmax

An activation function that outputs a probability distribution over multiple classes.

TensorBoard

A tool for visualizing and monitoring the training process of deep learning models.

Training Set

A subset of the data used to train the model.

Validation Set

A subset of the data used to evaluate the model's performance during training and tune hyperparameters.

Test Set

A subset of the data used to evaluate the final model's performance on unseen data.

Pretrained Model

A model that has been previously trained on a large dataset, providing a starting point for faster and more effective training.

Loss Function

A mathematical function that quantifies the difference between a model's predictions and the actual target values.

Metric

A human-interpretable measure used to evaluate the performance of a trained model, often different from the loss function.

Transfer Learning

The practice of leveraging a pretrained model for a new task, often involving adapting the model's architecture and fine-tuning its weights.

DataBlock

A blueprint for assembling datasets for deep learning in fastai, defining data input/output types, how to access data items, and more.

Convolution

A mathematical operation that forms the basis of convolutional layers in CNNs, extracting features from data.

Pooling Layers

Layers that downsample the feature maps produced by convolutional layers, reducing spatial dimensions while preserving information.

Learning Rate

A hyperparameter that controls the step size at each iteration while moving toward a minimum of a loss function.

Batch Gradient Descent

Refers to using the entire training dataset to compute the gradient and update parameters in a single step.

Mini-batch Gradient Descent

Involves randomly selecting a small subset of the training data to compute the gradient and update parameters.

Fine-tuning

A transfer learning technique where a pretrained model is trained on a new task to adapt it with additional epochs.

Cosine Similarity

A measure of similarity between two vectors, indicating how similar they are in terms of direction.

Latent Factors

Underlying characteristics that influence user preferences in collaborative filtering models.

Positive Feedback Loop

A process where the output of a system reinforces itself, potentially narrowing content recommendations.

Weight Decay

A regularization technique that discourages large weights in the model to prevent overfitting.

Principal Component Analysis (PCA)

A dimensionality reduction technique used to identify significant directions of variation in data.

Max Pooling

A pooling method that selects the maximum value from a defined pooling window, reducing dimensionality.

Average Pooling

A pooling method that calculates the average value within the pooling window.

DeBERTa

A transformer-based language model designed for various natural language processing tasks.

Autoregressive Model

A model that predicts future sequence values based on past observed values.

Masked Language Modeling

A training technique that involves hiding words in the input so that the model learns to predict them.

Permuted Language Modeling

A training technique that randomizes the order of words so the model learns to predict the original sequence.

Sequence Classification

Assigning a category or label to an entire sequence of text.

Stemmer

A tool that reduces words to their base or root form to simplify text data.

Special Tokens

Tokens added to text to provide specific instructions or information for the model.

Tokenization

The process of converting text into individual tokens or words for processing in machine learning.

Vision Transformers

A model that uses self-attention mechanisms for image recognition tasks.

Keras Functional API

A way to build complex neural network models flexibly by defining layers as functions and chaining them.

Sequential API

An API in Keras that allows adding layers to models in a simple, linear sequence.

Deep and Wide Networks

Neural networks that combine deep learning with wide network capabilities for better generalization.

Stochastic Gradient Descent (SGD)

An optimization algorithm that updates model weights incrementally based on small batches of data.

Feedback Loop

The process where the outcomes of a system feed back into that system, potentially reinforcing biases.

Hugging Face Repository

A platform that hosts pre-trained language models and tools for natural language processing tasks.

Multilingual Models

Models trained on multiple languages to handle and understand diverse language inputs.

Normalization

A preprocessing step that adjusts the values in the dataset to fall within a standard range.

Optimization

The process of adjusting the model's parameters to minimize the loss function during training.

Classification

The task of assigning labels to data points based on learned features from the dataset.

Clustering

A method of grouping data points based on similarity, often used in unsupervised learning.

Feature Extraction

The process of selecting and transforming raw data into informative features for model training.

Regularization Techniques

Methods employed to prevent overfitting in machine learning models, such as dropout or weight decay.

Hyperparameters

Parameters set before training a model, influencing the learning process and structure.

Embedding

A representation of high-dimensional data in a lower-dimensional space that captures relationships.

Neural Network

A computational model composed of interconnected nodes that processes data and generates outputs.

Training Process

The phase where the model learns patterns from data by minimizing a loss function over epochs.

Evaluation Metrics

Quantitative measures used to assess the performance of a machine learning model.

AutoML

Automated machine learning processes that facilitate the construction of models without extensive manual intervention.

Data Augmentation

Techniques to artificially expand the size of a training dataset by creating modified versions of existing data.

Image Recognition

The ability of a system to identify objects, people, places, or actions in images.

Natural Language Processing (NLP)

A field of artificial intelligence focused on the interaction between computers and human language.

Sentiment Analysis

The process of determining the emotional tone behind a series of words, used to analyze customer feedback.

Transfer Learning

A machine learning method where a model developed for one task is reused as the starting point for a model on a second task.

Precision

The ratio of true positive predictions to the total positive predictions made by a model.

Recall

The ratio of true positive predictions to the actual positive cases in the dataset.

F1 Score

A measure of a model's accuracy that considers both precision and recall to provide a balance.

Confusion Matrix

A matrix that summarizes the performance of a classification algorithm by showing true vs. predicted classifications.

Anomaly Detection

The identification of rare items, events, or observations that raise suspicions by differing significantly from the majority of the data.

Artificial Neural Network (ANN)

A network structure inspired by biological neural networks that processes inputs and produces outputs based on learned weights.

Feature Scaling

Methods used to normalize the range of independent variables or features of data.

Prototyping

The process of creating an initial model to test and iterate on an idea or product.