1) Given an accelerator with no memory to store the previous layer activations, Can you reuse the convolution output of a particular patch of the image for the next convolution operation? If so how, if not why? 2) Is it possible to reduce the multiple convolutions into a single convolution operation to speed up the execution?
Some basic software engineering question
General tech about C++
Why compilers do loop unrolling?
Coding Question asking me basic algorithms
1. Questions from the resume. 2. Vectorization 3. Difference between a structure and a class
I was being asked two coding questions, two design questions, and some math questions. One of the design questions is about designing the intermediate representations.
Describe projects mentioned on CV.
find how many warehouses are needed at least for a one-way road connecting major neighborhoods.
With respect to the coding question, I was asked a very tricky question related to topological sorting of nodes of a graph.
Viewing 91 - 100 interview questions