"Generative AI", Trustable-AI" and "Self-Supervised Learning"
Generative AI
- Variational Autoencoder.
Content:
- Part-1
- Basics of Autoencoders
- Basics of Variational Autoencoders
- Why Variational Autoencoder works like Generative Model but Traditional Autoencoder Not?
- Step-by-Step discussion on Variational Autoencoder.
- Part-2.
- Variational Autoencoder
- Variational Inference
- Computing - Cost function and other Mathematical formulations
Video Link:
Variational Autoencoder Part-1 (Direct Link: https://youtu.be/Mu7RoJHYqr4)
Variational Autoencoder Part-2 (Direct Link: https://youtu.be/bdPgByjoSH4)
- Generative Adversarial Network (GAN).
Content:
- Part-1
- Basics of GAN
- Function and Technique behind:
- Generator Model
- Discriminator Model
- GAN Architecture
- GAN Loss Function
- Part-2.
- Deep Convolutional GAN (DCGAN)
- GAN key issues in Detail
- Mode Collapse Problem
- Training Instability
- Overfitting Issues
- Convolutional Up sampling
- Convolutional Down sampling
- Part-3
- Step-By-Step Process to write the code for DCGAN.
- Detailed explanation behind each components of the code.
Video Link:
Generative Adversarial Network GAN Part-1 (Direct Link:https://youtu.be/TneKa3WHYO4)
Generative Adversarial Network GAN Part-2 (Direct Link: https://youtu.be/rWh9ov9mR_I?si=jcuJSVacr4CXuVpy)
Generative Adversarial Network GAN Part-3 (Direct Link: https://youtu.be/aBlvgN5w9sY?si=AjEP9jaira9vftNw)
- Wasserstein GAN (WGAN).
Content:
- Part-1
- Basics of Wasserstein GAN.
- Why we need Wasserstein Distance?
- Difference between – (a) KL-Divergence, Vs (b) Jensen-Shannon Divergence Vs (c) Wasserstein Distance
- Part-2.
- Wasserstein Distance.
- Wasserstein Distance through - (a) Kantrovich-Rubinstein Duality and (b) 1-Lipschitz Function
- Part-3
- Wasserstein GAN Architecture.
- Wasserstein GAN KERAs implementation
Video Link:
Wasserstein GAN Part-1(KL-Divergence Vs Jensen-Shannon Divergence Vs Wasserstein Distance) (Direct Link: https://youtu.be/4uEIMCa_BAo?si=6IP8qZogcQYokzjA)
Wasserstein GAN Part-2 (Wasserstein Distance - Details) (Direct Link: https://youtu.be/5hNgGSN8ScI?si=8QiUW0hksIYzgVfE)
Wasserstein GAN Part-3 (Architecture and Implementation) (Direct Link: https://youtu.be/m0oKu6u9X5o?si=Y1_yH4fmXDHj6fz6)
Self-Supervised Learning
- Deep Clustering.
Content:
- Basics of Self-Supervised Algorithm
- Details of Deep Clustering
- Details of Cost Functions used in the Deep Clustering Algorithms
Video Link:
Deep Clustering- Part-1 (A Self-Supervised Deep Learning Algorithm) (Direct Link: https://youtu.be/j9KmEpaLers )
Deep Clustering- Part-2 (A Self-Supervised Deep Learning Algorithm) (Direct Link: https://youtu.be/Ca0r0ZbeHxM )
- L2-Norm and Unit-Sphere for Contrastive/Self Supervised Learning.
Content:
- Basics of L2-Norm
- Basics of Unit-Sphere
- Why we use both in Contrastive /Self-Supervised Learning.
Video Link:
L2-Norm and Unit-Sphere for Contrastive/Self Supervised Learning (Direct Link: https://youtu.be/X6TcKMA_rNY )
3. Self-Supervised Online Clustering (Unsupervised Learning of Visual Features).
Content:
- Basics of Contrastive Learning
- Paper Discussion: “Unsupervised Learning of Visual Features by Contrastive Cluster Assignments”
Video Link:
Self-Supervised Online Clustering (Unsupervised Learning of Visual Features) (Direct Link: https://youtu.be/3z8L4jGK2FE )
Trustable AI
- Trustable AI -1.
Content:
- Basics of Trustable AI
- Uncertainty Estimation
- Aleatoric Uncertainty
- Epistemic Uncertainty
- Model Reliability Model Calibration