Kumar Desappan

Title: Innovations of Kumar Desappan

Introduction

Kumar Desappan is a notable inventor based in Bengaluru, India. He has made significant contributions to the field of technology, particularly in the area of data quantization for convolutional neural networks. With a total of four patents to his name, Desappan's work is recognized for its innovative approaches and practical applications.

Latest Patents

One of Kumar Desappan's latest patents is titled "Parametric power-of-2 clipping activations for quantization for convolutional neural networks." This patent describes a method for quantizing data for a convolutional neural network (CNN). The method involves receiving a set of data and quantizing it using a power-of-2 parametric activation (PACT2) function. The PACT2 function organizes the data into a histogram and discards a portion corresponding to the tail of the histogram, forming a remaining set of data. A clipping value is then determined by expanding this remaining set to the nearest power of two. This innovative approach allows for quantization either through post-training quantization or quantization-aware training, helping to achieve accuracy close to that of the corresponding floating-point model.

Another significant patent by Desappan is focused on "Methods and systems for masking multimedia data." This patent discloses several methods for masking multimedia data, where a prediction is performed for at least one multimedia data block based on various prediction modes. The method generates a residual multimedia data block associated with a region of interest (ROI) and performs quantization based on a variable quantization parameter (QP) value. This flexibility in the QP value allows for controlled masking of the ROI, showcasing Desappan's innovative approach to multimedia data processing.

Career Highlights

Kumar Desappan is currently employed at Texas Instruments Corporation, where he continues to develop and refine his innovative ideas. His work at Texas Instruments has allowed him to contribute to cutting-edge technology and collaborate with other talented professionals in the field.

Collaborations

Some of Kumar Desappan's coworkers include Soyeb Noormohammed Nagori and Manu Mathew. Their collaboration has likely fostered an environment of creativity and innovation, leading to advancements in their respective projects.

Conclusion

Kumar Desappan's contributions to technology through his patents demonstrate his commitment to innovation and excellence. His work in quantization methods for convolutional neural