I am currently a PhD student with Dr. Miguel Eckstein in the Psychological and Brain Sciences program at University of California, Santa Barbara. During my undergraduate days, I worked on control systems project aiming to improve training efficiency and reliability of EMG-controlled prostheses. This led me to develop interest in biological signals and neural computation. After graduation, I spent 3 years under the guidance of Dr. S.P. Arun at Indian Institute of Science Bangalore, studying various problems related to visual object recognition. Over the three years, I studied a variety of problems such as visual search asymmetries, face category learning and memorability, dissimilarity measures and their use in characterizing perceptual spaces, neuronal coding of object permanence, occlusion and shadows. During my PhD, I wish to acquire sufficient expertise in the technique of computational modelling. This will equip me with necessary tools to study visual cognition at 3 levels: behavioral, neuronal and computational. I believe that a complete solution to any problem in vision science must reconcile the observations/predictions from each of these three methods of data sampling.
- Neural coding of objects in higher visual areas: Photons impinging on the retina stimulate photo-sensitive cells known as Rods and Cons, which relay the information to the brain in the form of digitized electric signals known as action potentials. The signal is processed across several layers of the cortex (visual hierarchy) before we have the percept of meaningful vision. While the computations in lower levels of the visual hierarchy as quite well understood, we have no clear answer to the question of what computations happen in higher levels of visual processing. This is partly because a steep increase in complexity of the neuronal code in higher visual areas owing to multiple layers of non-linear processing. Other reasons include increasing receptive field sizes, increasing top down influence of attention, sparsity and distributed nature of neural code. I use the following techniques to study neuronal coding in higher visual areas:
- Characterizing tuning function of neurons to a feature: Neuronal responses from cells (or BOLD signals from voxels in fMRI data) to specially crafted stimuli that differ only across a certain feature are measured and the tuning function is studied to check if a certain feature is being coded for.
- Neural correlates for a certain behavioral phenomenon: Behavioral studies often yield insights into how the brain might be using visual information. An important technique to understand neural computations in higher areas is to trace the evolution of this kind of information in neuronal code, as visual processing happens along the visual pathway.
- Computational models of spiking: Neurons in the brain are nodes in a complex network of neural circuitry. The current belief is that all the complex semantic information from visual stimuli is extracted through mutual excitation/inhibition between neurons in different layers of the brain. Another way to study neuronal coding is to reverse engineer a computational model that produces the required format of neural code (as verified through experiments) from the input stimulus data.
- Face processing: Faces are a class of commonly encountered complex stimuli. They are complex because fine differences in face features make a big difference in face identity. Humans are naturally good at face recognition, and even infants are known to accurately identify faces of their mothers. The basic puzzle of faces processing is to find out what features on a face are used by the brain for face recognition, and how. While face recognition algorithms have been studied extensively from a computer vision perspective, mechanisms of face processing in humans are relatively less understood. Two techniques I use to study human face recognition are:
- Studying behavioral performance as a function of available input: These techniques aim to characterize face recognition performance as function of the amount of information available to the subject from a face. If the occlusion/modification of a certain part of the face affects face recognition significantly, the information in that part is probably important for face recognition.
- Characterization of face space: Each face could be visualized to have a unique coordinate in a space where the dimensions represent features on the face the brain cares about. The process of recognition then can be likened to successfully resolving the location of a face in this space with respect to other members.
- Visual category learning: One striking quality of human behavior is ability to learn generalized rules based on repeating patterns observed in nature. One of the basic questions concerning this process is how these patterns are detected and adopted. What is the role of memory in learning? Is there an interaction between memory and learning? These questions can be studied through the use of simple psychophysics experiments, where subjects undergo training in a simple categorization task, while their memory is simultaneously tested. Any interactions found between learning rates and memory would provide clues regarding how brain makes a shift in strategy between rote memories of stimuli to the usage of patterns to accomplish the task. Obtaining fMRI data during category learning can yield important clues regarding what areas in the brain are implicated in this shift of strategy.
H. Katti, N.C. Puneeth, and S.P. Arun, Competitive Interactions between rule and association learning in face categorization, 824.11/JJ6, Neuroscience 2014, Washington D.C., Society for Neuroscience, 2014. Abstract available online.
- In my free time, I enjoy hiking, cooking, playing ping pong and the piano.
- I hail from the historic city of Hyderabad, a cultural center in India, famous for its biryani (a spicy rice dish).
- I love to discuss philosophy, science and technology. If you have something interesting to share with me, feel free to reach out at chakravarthula[at]psych[dot]ucsb.edu.
Fax: (805) 893-4303
Department of Psychology
University of California, Santa Barbara
Santa Barbara, CA 93106-9660