Project 2: Disha Jindal

[DishaJindal]

• Repo Link
• Stream Compaction
  • CPU Scan & Stream Compaction
  • Naive GPU Scan Algorithm
  • Work-Efficient GPU Scan & Stream Compaction
  • Thrust's Implementation
  • Performance Analysis
    [Extra Credit]
  • Optimized Work-Efficient GPU Scan & Stream Compaction: Since, the work efficient GPU approach has unnecessary number of live threads, I am making sure that we only have as many number of threads as required.
  • Radix Sort: Implemented Radix Sort using the optimized work-efficient stream compaction
• Character Recognition
  • Multi Layer Perceptron with Input, Hidden and Output Layers
  • Trained and Tested on both XOR and Character Recognition Datasets
  • Tried Softmax and ReLU Activations
  • Performance Analysis w.r.t. number of nodes in hidden layer and learning rate