Bioinspired random projections for robust, sparse classification

@article{Davies2022BioinspiredRP,
  title={Bioinspired random projections for robust, sparse classification},
  author={Bryn Davies and Nina Dekoninck Bruhin},
  journal={ArXiv},
  year={2022},
  volume={abs/2206.09222}
}
Inspired by the use of random projections in biological sensing systems, we present a new algorithm for processing data in classification problems. This is based on observations of the human brain and the fruit fly’s olfactory system and involves randomly projecting data into a space of greatly increased dimension before applying a cap operation to truncate the smaller entries. This leads to a simple algorithm that is very computationally efficient and can be used to either give a sparse… 

Figures from this paper

References

SHOWING 1-10 OF 40 REFERENCES

An algorithmic theory of learning: Robust concepts and random projection

  • R. ArriagaS. Vempala
  • Computer Science
    40th Annual Symposium on Foundations of Computer Science (Cat. No.99CB37039)
  • 1999
TLDR
This work provides a novel algorithmic analysis via a model of robust concept learning (closely related to “margin classifiers”), and shows that a relatively small number of examples are sufficient to learn rich concept classes.

Random Projection in the Brain and Computation with Assemblies of Neurons

TLDR
It is proved that the operation of assembly projection converges with high probability, over the randomness of synaptic connectivity, even if plasticity is relatively small (previous proofs relied on high plas19 ticity).

Visual Categorization with Random Projection

TLDR
It is found that drastic reduction in stimulus complexity via random projection does not degrade performance in categorization tasks by either humans or simple neural networks, and human accuracy and neural network accuracy are remarkably correlated, even at the level of individual stimuli.

Multi-scale classification for electro-sensing

TLDR
This paper introduces premier and innovative (real-time) multi-scale method for target classification in electro-sensing based on a family of transform-invariant shape descriptors computed from generalized polarization tensors reconstructed at multiple scales.

A neural algorithm for a fundamental computing problem

TLDR
On the basis of the algorithm that flies use to tag an odor and categorize similar ones, the authors generated a new solution to the nearest-neighbor search problem that underlies tasks such as searching for similar images on the web.

A statistical property of fly odor responses is conserved across odors

  • C. Stevens
  • Biology
    Proceedings of the National Academy of Sciences
  • 2016
TLDR
The fly odor code is characterized by showing it is maximum entropy; what this means is that the response to all odors is statistically the same, with the mean rate dependent on the odor concentration.

A circuit supporting concentration-invariant odor perception in Drosophila

TLDR
The results show that increasing odor concentrations induce progressive activation of concentration-tuned olfactory sensory neurons and concomitant recruitment of inhibitory local interneurons, and it is proposed that the interplay of combinatorial OSN input and localinterneuron activation allows animals to remain sensitive to odors across a large range of stimulus intensities.

Face recognition experiments with random projection

TLDR
The experimental results illustrate that although RP represents faces in a random, low-dimensional subspace, its overall performance is comparable to that of PCA while having lower computational requirements and being data independent.

Imaging a Population Code for Odor Identity in the Drosophila Mushroom Body

TLDR
This work uses two-photon Ca2+ imaging to record odor-evoked responses from >100 neurons simultaneously in the Drosophila mushroom body and demonstrates quantitatively that MB population responses contain substantial information on odor identity.

Olfactory Information Processing in Drosophila