This is a static page that lists all courseware - lecture topics, slides, and reading material. Details of a specific run of the course (eg. evaluation pattern and assignments) are in Moodle or Google Classroom.

Please drop me a line in case you find a missing or broken link.

Pre-requisites

---

• Computer Organization (CS2600) / Computer Architecture (CS6600)

• Pattern Recognition and Machine Learning (CS5691) / Machine Learning for Computer Vision (EE5177)

Usually students find this course easier to follow if they have done Fundamentals in Deep Learning (CS6910).

Courseware

---

Contents	Lecture Slides	Additional Slides	Reading
Logistics	Lecture	-	-
Overview	Lecture	-	A Berkeley View | MLSys | AI for systems, SIGARCH blog | Deep Learning Hardware | Systems & Machine Learning
Systems thinking	Lecture	Systems thinking | Principles of SW Design | Chaos | Intel SGX	Chapter 1 from "Principles of Computer System Design" | No silver bullet in SE | Future Computing Architectures | A Vision for the Future of Systems
Key Ideas in ML	Lecture	Two cultures of stats | XGBoost	Few useful things to know in ML | Thinking in higher dimension | Statistical Modeling: The Two Cultures
A quick introduction to Deep Learning	Lecture	-	Visualising convolution Popular DNNs and Datasets | DL: An overview | MIT DL Basics |
Age of Accelerators	Lecture Hardware Architectures for DNN | DL: A systems perspective	Slides from Turing Award
MLPerf	Lecture	-	MLPerf Inference Benchmark | The Tail at Scale | Weights and Bias Tool |
Hardware for DL	Lecture 1 | Lecture 2 | Lecture 3 | Lecture 4	Survey | Tensor Processing Unit | Gemmini Paper | SCALE-Sim paper	Efficient Processing of DNNs: A tutorial and Survey Plotting a Course to continued Moore's Law: Partha Ranganathan (First 22 mins)
Systolic Algorithms	Lecture	Convolution with Systolic Array	-
DL Frameworks	Lecture	TensorFlow | PyTorch | Janus | TensorFlow Eager | PyTorch Geometric	CS231n Lecture 8 : Deep Learning Software TensorFlow OSDI
Automatic Differentation	Lecture	-	Automatic differentiation in ML: Where we are and where we should be going Lecture on Backpropagation Lecture on Auto Diff Halide Paper
History of Programming Languages	Lecture	Julia	-
DL Compilers	Lecture	XLA | TASO | TVM	TVM paper
Cloud Deployment and Efficient Inference	Lecture	MLOps | MLFlow | FPGA DNN accelerators | TF Lite | Nvidia TensorRT |
NAS	Lecture	DARTS | OFA | Attentive NAS |	DARTS paper
Miscellaneous		Lottery Ticket Hypothesis | NeuroEvolution | ML in EDA | Memristors for Inference	Lottery Ticket Hypothesis paper | Deep Neuroevolution paper | ML applications in Physical Design | PUMA paper