Yijiang River Dong, Hongzhou Lin, Mikhail Belkin, Ramon Huerta, Ivan Vulić. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

The broad receptive field of the olfactory receptors constitutes the basis of a combinatorial code that allows animals to detect and discriminate many more odorants than the actual number of receptor types that they express. One drawback is that high odor concentrations recruit lower affinity receptors which can lead to the perception of qualitatively different odors. Here we addressed the contribution that signal-processing in the antennal lobe makes to reduce concentration dependence in odor representation. By means of calcium imaging and pharmacological approach we describe the contribution that GABA receptors play in terms of the amplitude and temporal profiles of the signals that convey odor information from the antennal lobes to higher brain centers. We found that GABA reduces the amplitude of odor elicited signals and the number of glomeruli that are recruited in an odor-concentration-dependent manner. Blocking GABA receptors decreases the correlation among glomerular activity patterns elicited by different concentrations of the same odor. In addition, we built a realistic mathematical model of the antennal lobe that was used to test the viability of the proposed mechanisms and to evaluate the processing properties of the AL network under conditions that cannot be achieved in physiology experiments. Interestingly, even though based on a rather simple topology and cell interactions solely mediated by GABAergic lateral inhibitions, the AL model reproduced key features of the AL response upon different odor concentrations and provides plausible solutions for concentration invariant recognition of odors by artificial sensors.

MOTIVATION: Despite RNA-seq reads provide quality scores that represent the probability of calling a correct base, these values are not probabilistically integrated in most alignment algorithms. Based on the quality scores of the reads, we propose to calculate a lower bound of the probability of alignment of any fast alignment algorithm that generates SAM files. This bound is called Fast Bayesian Bound (FBB) and serves as a canonical reference to compare alignment results across different algorithms. This Bayesian Bound intends to provide additional support to the current state-of-the-art aligners, not to replace them. RESULTS: We propose a feasible Bayesian bound that uses quality scores of the reads to align them to a genome of reference. Two theorems are provided to efficiently calculate the Bayesian bound that under some conditions becomes the equality. The algorithm reads the SAM files generated by the alignment algorithms using multiple command option values. The program options are mapped into the FBB reference values, and all the aligners can be compared respect to the same accuracy values provided by the FBB. Stranded paired read RNA-seq data was used for evaluation purposes. The errors of the alignments can be calculated based on the information contained in the distance between the pairs given by Theorem 2, and the alignments to the incorrect strand. Most of the algorithms (Bowtie, Bowtie 2, SHRiMP2, Soap 2, Novoalign) provide similar results with subtle variations. AVAILABILITY AND IMPLEMENTATION: Current version of the FBB software is provided at https://bitbucket.org/irenerodriguez/fbb CONTACT: [email protected] information: Supplementary data are available at Bioinformatics online.

Gain control is essential for the proper function of any sensory system. However, the precise mechanisms for achieving effective gain control in the brain are unknown. Based on our understanding of the existence and strength of connections in the insect olfactory system, we analyze the conditions that lead to controlled gain in a randomly connected network of excitatory and inhibitory neurons. We consider two scenarios for the variation of input into the system. In the first case, the intensity of the sensory input controls the input currents to a fixed proportion of neurons of the excitatory and inhibitory populations. In the second case, increasing intensity of the sensory stimulus will both, recruit an increasing number of neurons that receive input and change the input current that they receive. Using a mean field approximation for the network activity we derive relationships between the parameters of the network that ensure that the overall level of activity of the excitatory population remains unchanged for increasing intensity of the external stimulation. We find that, first, the main parameters that regulate network gain are the probabilities of connections from the inhibitory population to the excitatory population and of the connections within the inhibitory population. Second, we show that strict gain control is not achievable in a random network in the second case, when the input recruits an increasing number of neurons. Finally, we confirm that the gain control conditions derived from the mean field approximation are valid in simulations of firing rate models and Hodgkin-Huxley conductance based models.

The role of inhibition is investigated in a multiclass support vector machine formalism inspired by the brain structure of insects. The so-called mushroom bodies have a set of output neurons, or classification functions, that compete with each other to encode a particular input. Strongly active output neurons depress or inhibit the remaining outputs without knowing which is correct or incorrect. Accordingly, we propose to use a classification function that embodies unselective inhibition and train it in the large margin classifier framework. Inhibition leads to more robust classifiers in the sense that they perform better on larger areas of appropriate hyperparameters when assessed with leave-one-out strategies. We also show that the classifier with inhibition is a tight bound to probabilistic exponential models and is Bayes consistent for 3-class problems. These properties make this approach useful for data sets with a limited number of labeled examples. For larger data sets, there is no significant comparative advantage to other multiclass SVM approaches.

Motivated by the insect olfactory system, which resolves both the identity and the quantity of a nectar in parallel based on the same sensory cue, we address the problem of Volatile Organic Compound (VOC) classification and regression in a unified setting. We derive a maximum margin formulation for minimizing the empirical regression error and the classification error jointly, and then call the sequential minimal optimization procedure for solution. The solution yields a pool of support vectors that achieves both tasks almost equally accurately as individual performances of a support vector machine classifier and a support vector regressor designed independently. We investigate empirically the advantages and inconveniences of handling these two problems under a single formulation for odor identification and quantification. We demonstrate the method on an extensive dataset acquired by an array metal-oxide sensors for five VOC identities and a wide range of concentrations.

We propose a model for pattern recognition in the insect brain. Departing from a well-known body of knowledge about the insect brain, we investigate which of the potentially present features may be useful to learn input patterns rapidly and in a stable manner. The plasticity underlying pattern recognition is situated in the insect mushroom bodies and requires an error signal to associate the stimulus with a proper response. As a proof of concept, we used our model insect brain to classify the well-known MNIST database of handwritten digits, a popular benchmark for classifiers. We show that the structural organization of the insect brain appears to be suitable for both fast learning of new stimuli and reasonable performance in stationary conditions. Furthermore, it is extremely robust to damage to the brain structures involved in sensory processing. Finally, we suggest that spatiotemporal dynamics can improve the level of confidence in a classification decision. The proposed approach allows testing the effect of hypothesized mechanisms rather than speculating on their benefit for system performance or confidence in its responses.

The speed and accuracy of odor recognition in insects can hardly be resolved by the raw descriptors provided by olfactory receptors alone due to their slow time constant and high variability. The animal overcomes these barriers by means of the antennal lobe (AL) dynamics, which consolidates the classificatory information in receptor signal with a spatiotemporal code that is enriched in odor sensitivity, particularly in its transient. Inspired by this fact, we propose an easily implementable AL-like network and show that it significantly expedites and enhances the identification of odors from slow and noisy artificial polymer sensor responses. The device owes its efficiency to two intrinsic mechanisms: inhibition (which triggers a competition) and integration (due to the dynamical nature of the network). The former functions as a sharpening filter extracting the features of receptor signal that favor odor separation, whereas the latter implements a working memory by accumulating the extracted features in trajectories. This cooperation boosts the odor specificity during the receptor transient, which is essential for fast odor recognition.

The odor transduction process has a large time constant and is susceptible to various types of noise. Therefore, the olfactory code at the sensor/receptor level is in general a slow and highly variable indicator of the input odor in both natural and artificial situations. Insects overcome this problem by using a neuronal device in their Antennal Lobe (AL), which transforms the identity code of olfactory receptors to a spatio-temporal code. This transformation improves the decision of the Mushroom Bodies (MBs), the subsequent classifier, in both speed and accuracy.Here we propose a rate model based on two intrinsic mechanisms in the insect AL, namely integration and inhibition. Then we present a MB classifier model that resembles the sparse and random structure of insect MB. A local Hebbian learning procedure governs the plasticity in the model. These formulations not only help to understand the signal conditioning and classification methods of insect olfactory systems, but also can be leveraged in synthetic problems. Among them, we consider here the discrimination of odor mixtures from pure odors. We show on a set of records from metal-oxide gas sensors that the cascade of these two new models facilitates fast and accurate discrimination of even highly imbalanced mixtures from pure odors.

The idea that cognitive activity can be understood using nonlinear dynamics has been intensively discussed at length for the last 15 years. One of the popular points of view is that metastable states play a key role in the execution of cognitive functions. Experimental and modeling studies suggest that most of these functions are the result of transient activity of large-scale brain networks in the presence of noise. Such transients may consist of a sequential switching between different metastable cognitive states. The main problem faced when using dynamical theory to describe transient cognitive processes is the fundamental contradiction between reproducibility and flexibility of transient behavior. In this paper, we propose a theoretical description of transient cognitive dynamics based on the interaction of functionally dependent metastable cognitive states. The mathematical image of such transient activity is a stable heteroclinic channel, i.e., a set of trajectories in the vicinity of a heteroclinic skeleton that consists of saddles and unstable separatrices that connect their surroundings. We suggest a basic mathematical model, a strongly dissipative dynamical system, and formulate the conditions for the robustness and reproducibility of cognitive transients that satisfy the competing requirements for stability and flexibility. Based on this approach, we describe here an effective solution for the problem of sequential decision making, represented as a fixed time game: a player takes sequential actions in a changing noisy environment so as to maximize a cumulative reward. As we predict and verify in computer simulations, noise plays an important role in optimizing the gain.

A study of a general central pattern generator (CPG) is carried out by means of a measure of the gain of information between the number of available topology configurations and the output rhythmic activity. The neurons of the CPG are chaotic Hindmarsh-Rose models that cooperate dynamically to generate either chaotic or regular spatiotemporal patterns. These model neurons are implemented by computer simulations and electronic circuits. Out of a random pool of input configurations, a small subset of them maximizes the gain of information. Two important characteristics of this subset are emphasized: (1) the most regular output activities are chosen, and (2) none of the selected input configurations are networks with open topology. These two principles are observed in living CPGs as well as in model CPGs that are the most efficient in controlling mechanical tasks, and they are evidence that the information-theoretical analysis can be an invaluable tool in searching for general properties of CPGs.

We propose a theoretical framework for odor classification in the olfactory system of insects. The classification task is accomplished in two steps. The first is a transformation from the antennal lobe to the intrinsic Kenyon cells in the mushroom body. This transformation into a higher-dimensional space is an injective function and can be implemented without any type of learning at the synaptic connections. In the second step, the encoded odors in the intrinsic Kenyon cells are linearly classified in the mushroom body lobes. The neurons that perform this linear classification are equivalent to hyperplanes whose connections are tuned by local Hebbian learning and by competition due to mutual inhibition. We calculate the range of values of activity and size of the network required to achieve efficient classification within this scheme in insect olfaction. We are able to demonstrate that biologically plausible control mechanisms can accomplish efficient classification of odors.

Experimental observations of the intracellular recorded electrical activity in individual neurons show that the temporal behavior is often chaotic. We discuss both our own observations on a cell from the stomatogastric central pattern generator of lobster and earlier observations in other cells. In this paper we work with models with chaotic neurons, building on models by Hindmarsh and Rose for bursting, spiking activity in neurons. The key feature of these simplified models of neurons is the presence of coupled slow and fast subsystems. We analyze the model neurons using the same tools employed in the analysis of our experimental data. We couple two model neurons both electrotonically and electrochemically in inhibitory and excitatory fashions. In each of these cases, we demonstrate that the model neurons can synchronize in phase and out of phase depending on the strength of the coupling. For normal synaptic coupling, we have a time delay between the action of one neuron and the response of the other. We also analyze how the synchronization depends on this delay. A rich spectrum of synchronized behaviors is possible for electrically coupled neurons and for inhibitory coupling between neurons. In synchronous neurons one typically sees chaotic motion of the coupled neurons. Excitatory coupling produces essentially periodic voltage trajectories, which are also synchronized. We display and discuss these synchronized behaviors using two "distance" measures of the synchronization.

We present two methods for the prediction of coupled time series. The first one is based on modeling the series by a dynamic system with a polynomial format. This method can be formulated in terms of learning in a recurrent network, for which we give a computationally effective algorithm. The second method is a purely feedforward σ-π network procedure whose architecture derives from the recurrence relations for the derivatives of the trajectories of a Ricatti format dynamic system. It can also be used for the modeling of discrete series in terms of nonlinear mappings. Both methods have been tested successfully against chaotic series.