Neural Networks and Deep Learning Research
Neural Networks and Deep Learning Research
Columbia University course ECBM E6040
Zoran Kostic, Ph.D., Dipl. Ing., Professor of Professional Practice, zk2172(at)columbia.edu
Electrical Engineering Department, Data Sciences Institute, Columbia University in the City of New York
Course in a nutshell:
Advanced theory and practice of Deep Learning. Applications and projects.
Description: Advanced (second) course on Deep Learning
Bulletin Description: Regularized autoencoders, sparse coding and predictive sparse decomposition, denoising autoencoders, representation learning, manifold perspective on representation learning, structured probabilistic models for deep learning, Monte Carlo methods, training and evaluating models with intractable partition functions, restricted Boltzmann machines, approximate inference, deep belief networks, deep learning in speech and object recognition.
Time:
Spring 2020: Wednesdays 8:00am to 10:00 am
Spring 2019: Wednesdays 8:00am to 10:00 am
Spring 2018: Mondays 8:00am to 10:00 am.
Prerequisites:
Required prerequisites: knowledge of linear algebra, probability and statistics, programming, machine learning, first course in deep learning.
Prerequisite courses: ECBM E4040 or similar
Content
Convolutional and recurrent neural networks.
Analytical study and software design.
Sevearal assignments in Python and in PyTorch
Significant project.
Pursuing deeper exploration of deep learning.
Organization
Lectures:
Presentation of material by instructors and guest lecturers
Student Presentations
Every student contributes several presentations on the subject of interest
Assignments:
Combination of analytical and programming assignments
Projects:
Team-based
Students with complementary backgrounds
Significant design
Reports and presentations to Columbia and NYC community
Best projects could qualify for publications and/or funding
Project Areas
Smart cities
Medical
Autonomous vehicles
Environmental
Physical data analytics
Finance
Books, Tools and Resources
BOOKS:
2020 software platform:
PyTorch as the main framework, Google TensorFlow, Google Cloud, Python, bitbucket
2020 software:
2020 Spring Projects
Personality Detection
Edge Detection
DQN for game AI
Lane Detection
IOU, Lovasz-Softmax
Medical Imaging, Self-supervised, Object Detection
Training a DeeqQ Agent to control traffic lights
Future Vehicle Localization
2019 Projects
AlphaXENAS - Efficient neural architecture search by using reinforcement learning and MCTS
Quantitative Comparison of NLP Embeddings
Distruptive Optimization using Gaussian Gradients
Neural Segmentation of Podcast Content
Learning to generate captions for images
MAD-GAN: Multivariate Anomaly Detection for Time Series Data with Generative Adversarial Networks
Meta Learning
3D Object tracking
Recurrent Neural Networks with Novel Regularization Mechanism
MRI Super-resolution
Semi-supervised Variational Autoencoder
2018 Projects
SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient
Neural Image Caption Generation with Visual Attention
To Better Understanding of Arts with Deep learning
DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks
Dialogue Generation with Reinforcement Learning
Dynamic Routing Between Capsules
Performance Profiling for Deep Learning
Generative Adversarial Text to Image Synthesis
Music Separation using Image Segmentation Networks (U-Net)
Real time object detection using Faster RCNN
A Neural Algorithm of Artistic Style
Course sponsored by equipment and financial contributions of:
NVidia GPU Education Center, Google Cloud, IBM Bluemix, AWS Educate, Atmel, Broadcom (Wiced platform); Intel (Edison IoT platform), Silicon Labs.