Knowledge Graph — Coursera Notes › Academic disciplines › Computer Science / Information Technology › Artificial Intelligence

Deep Learning

concept · part of Artificial Intelligence

A subset of ML using neural networks capable of processing complex inputs. Deep learning uses artificial neural networks inspired by the brain to learn from vast amounts of data, enabling tasks like image recognition, natural language processing, and game playing.

Foundational references include Goodfellow et al.'s 'Deep Learning' and LeCun et al.'s 2015 Nature review; key techniques are optimization, backpropagation, and architectures like CNNs and RNNs.

Inside Deep Learning (18)

• TensorFlow — A deep learning framework chosen for production scalability and tooling, used by Novartis for drug discovery.
• PyTorch — A deep learning framework chosen for dynamic graphs and Pythonic integration, used by Walmart for recommendation systems.
• Activation Functions — Activation functions determine the output of a neural network node, introducing non-linearity into the model.
• CNN — A convolutional neural network (CNN) is a neural network architecture with convolutional layers followed by pooling and fully connected layers, suited for image classification.
• Transformer — Underlying architecture of modern LLMs
• Caffe — A fast CNN training framework with a Model Zoo and low memory overhead, but specialized for CNNs and now effectively defunct.
• Model Training — The neural network is trained for 50 epochs with a batch size of 32.
• ReLU — An activation function that introduces non-linearity by outputting the input if positive, else zero.
• Apache MXNet — A hybrid symbolic+imperative framework with distributed scalability and multi-language support, but has a small community and low adoption.
• Deep learning for recommendations — Netflix employs deep learning neural networks to process vast amounts of data and improve recommendation accuracy.
• Dense — A Dense layer, also known as a fully connected layer, is a fundamental building block in neural networks where every neuron in the layer is connected to every neuron in the previous layer.
• Dropout — A regularization technique that randomly zeroes activations during training to prevent overfitting.
• LSTM — A variant of RNN that captures long-term dependencies in sequences.
• OpenVINO — Intel's hardware-specific optimization for hardware-accelerated inference on Intel CPUs, GPUs, and VPUs.
• RNN — A neural network architecture for sequential data where input order matters, processing tokens one by one.
• Softmax — An activation function that converts logits into a probability distribution over multiple classes.
• TensorRT — TensorRT is NVIDIA's high-performance deep learning inference optimizer and runtime library, designed to deliver low latency and high throughput for production deployments.
• Training loop — TensorFlow's fit method automates the training loop including batching, loss computation, and backpropagation.

Connections

• Related to Generative AI
• Uses TensorFlow
• Uses PyTorch
• Used for TensorFlow
• Used for PyTorch
• Related to Generative AI
• Related to Transformer
• Related to PyTorch
• Related to TensorFlow
• Related to CNN
• Related to ReLU
• Builds on Classification

This is the text view of an interactive 3D knowledge graph — open this page with JavaScript enabled to explore it visually.