logo

ELEC 515 Embedded Machine Learning - Schedule

Schedule Notes:

  • Section 1: Introduction of commonly used ML algorithms

  • Section 2: State-of-the-art algorithmic techniques of embedded ML

  • Section 3: State-of-the-art ML accelerator architectures and chips

Date Topic
Section 1: Introduction of commonly used ML algorithms
8/26/2019 (Mon.) Background, Overview, and Logistics
8/28/2019 (Wed.) DNN Inference and Training; Introduction to Google Colab
9/2/2019 (Mon.) Labor Day (No Scheduled Classes)
9/4/2019 (Wed.) Introduction to State-of-the-art DNNs
9/9/2019 (Mon.) Datasets, and Benchmarking Metrics; Deep Learning Software
9/11/2019 (Wed.) Decision Tree; Random Forests
9/16/2019 (Mon.) AdaBoost; Support Vector Machine Learning
9/18/2019 (Wed.) Generative Adversarial Network
9/23/2019 (Mon.) Reinforcement Learning
9/25/2019 (Wed.) Recurrent Neural Network; kNN, k-Means
9/30/2019 (Mon.) Summary of Section I
Section 2: State-of-the-art algorithmic techniques of embedded ML
10/2/2019 (Wed.) How to Use Embedded Machine Learning Tools: Mobiles and FPGAs
(Invited Speaker: Yue Wang and Pengfei Xu from Rice)
10/7/2019 (Mon.) Quantization Basics, and a Case Study: HAQ: Hardware-Aware Automated Quantization with Mixed Precision (CVPR 2019)
(Invited Speaker: Zhijian Liu from MIT)
10/9/2019 (Wed.) Quantized DNNs, and its Case Study: 8-bit Floating Point Training (NeurIPS 2018)
(Invited Speaker: Dr. Naigang Wang from IBM)
10/14/2019 (Mon.) Midterm Recess (No Scheduled Classes)
10/16/2019 (Wed.) Unstructured and Structured Pruning, and its Case Study: To Prune, or Not to Prune (ICLR 2018)
10/21/2019 (Mon.) Compression vis Early Exiting, Prediction, and its Case Studies.
10/23/2019 (Wed.) Compression via Parameter Sharing, Indexing, and its Case Study: Deep k-Means (ICML 2018)
10/28/2019 (Mon.) Regularization and Case Study of Unstructured Pruning: Learning Structured Sparsity in DNNs (NIPS 2016)
(Invited Speaker: Wei Wen from Duke)
10/30/2019 (Wed.) Compression via Transferred/Compact Convolutional Filters, and its Case Study: SqueezeNet ;
Compression via Knowledge Distallastion, and its Case Study: Distilling the Knowledge in a Neural Network
11/4/2019 (Mon.) Compression via Neural Architecture Search, and its Case Study: FBNet
(Invited Speaker: Dr. Bichen Wu from Berkeley)
Section 3: State-of-the-art ML accelerator architectures and chips
11/6/2019 (Wed.) Overview of DNN Acelerator Development, and DNN Accelerator Simulators/Chip-predictors
11/11/2019 (Mon.) Algorithm to Architecture Mapping I
11/13/2019 (Wed.) Algorithm to Architecture Mapping II
11/18/2019 (Mon.) Pipelined and Parallelled Processing as Low Power Techniques; Near Threshold Computing
11/20/2019 (Wed.) Energy-Delay TradeOff; DNN Accelerator Case Study (Eyeriss)
11/25/2019 (Mon.) DNN Accelerator Case Study (Google TPU, IBM TrueNorth, and DianNao series)
11/27/2019 (Wed.) Analog Processing Basics; DNN Accelerator Case Study: Towards artificial general intelligence with hybrid Tianjic chip architecture
(Invited Speaker: Dr. Lei Deng from UCSB)
12/2/2019 (Mon.) no class
12/4/2019 (Wed.) Processing-in-Memory Basics; DNN Accelerator Case Study: CirCNN: Accelerating and Compressing Deep Neural Networks Using Block-CirculantWeight Matrices
(Invited Speaker: Dr. Yanzhi Wang from Northeastern University)
12/5/2019 (Thur.) Course Project Presentation (15 mins Pres + 5 mins Q&A)