Lensa ML
Track 3 of 6

Deep Learning

Neurons and backprop, training deep networks, the main architectures, and computer vision.

You know classic ML and want to understand neural networks. 4 paths · 25 explainers.

Path 01 · 8 explainers

Neural Networks from Zero

From a single neuron to forward propagation, loss, and backprop

  1. 1Neural Networks
  2. 2The Perceptron
  3. 3Forward Propagation
  4. 4Activation Functions
  5. 5Bias & Decision Boundaries
  6. 6Softmax, Logits & Temperature
  7. 7Loss Functions
  8. 8Backpropagation
Path 02 · 7 explainers

Training Deep Networks

Optimizers, schedules, initialization, normalization, and keeping training stable

  1. 1Optimizers
  2. 2Learning Rate Scheduling
  3. 3Weight Initialization
  4. 4Batch Normalization
  5. 5Vanishing & Exploding Gradients
  6. 6Overfitting & Underfitting
  7. 7Regularization
Path 03 · 6 explainers

Architectures

CNNs, recurrent nets, autoencoders, GANs, and skip connections

  1. 1Convolutional Neural Networks
  2. 2Recurrent Neural Networks
  3. 3LSTMs
  4. 4Autoencoders & VAEs
  5. 5GANs
  6. 6Skip Connections & ResNets
Path 04 · 4 explainers

Computer Vision

Pooling, detection, segmentation, and data augmentation

  1. 1Pooling & Downsampling
  2. 2Object Detection
  3. 3Image Segmentation
  4. 4Data Augmentation