Lifetime learning, or the change (or acquisition) of behaviors during a lifetime, based on experience, is a hallmark of living organisms. Multiple mechanisms may be involved, but biological neural circuits have repeatedly demonstrated a vital role in the learning process. These neural circuits are recurrent, dynamic, and non-linear and models of neural circuits employed in neuroscience and neuroethology tend to involve, accordingly, continuous-time, non-linear, and recurrently interconnected components. Currently, the main approach for finding configurations of dynamical recurrent neural networks that demonstrate behaviors of interest is using stochastic search techniques, such as evolutionary algorithms. In an evolutionary algorithm, these dynamic recurrent neural networks are evolved to perform the behavior over multiple generations, through selection, inheritance, and mutation, across a population of solutions. Although, these systems can be evolved to exhibit lifetime learning behavior, there are no explicit rules built into these dynamic recurrent neural networks that facilitate learning during their lifetime (e.g., reward signals). In this work, we examine a biologically plausible lifetime learning mechanism for dynamical recurrent neural networks. We focus on a recently proposed reinforcement learning mechanism inspired by neuromodulatory reward signals and ongoing fluctuations in synaptic strengths. Specifically, we extend one of the best-studied and most-commonly used dynamic recurrent neural networks to incorporate the reinforcement learning mechanism. First, we demonstrate that this extended dynamical system (model and learning mechanism) can autonomously learn to perform a central pattern generation task. Second, we compare the robustness and efficiency of the reinforcement learning rules in relation to two baseline models, a random walk and a hill-climbing walk through parameter space. Third, we systematically study the effect of the different meta-parameters of the learning mechanism on the behavioral learning performance. Finally, we report on preliminary results exploring the generality and scalability of this learning mechanism for dynamical neural networks as well as directions for future work.
more »
« less
A unified theory for the origin of grid cells through the lens of pattern formation.
Grid cells in the brain fire in strikingly regular hexagonal patterns across space. There are currently two seemingly unrelated frameworks for understanding these patterns. Mechanistic models account for hexagonal firing fields as the result of pattern-forming dynamics in a recurrent neural network with hand-tuned center-surround connectivity. Normative models specify a neural architecture, a learning rule, and a navigational task, and observe that grid-like firing fields emerge due to the constraints of solving this task. Here we provide an analytic theory that unifies the two perspectives by casting the learning dynamics of neural networks trained on navigational tasks as a pattern forming dynamical system. This theory provides insight into the optimal solutions of diverse formulations of the normative task, and shows that symmetries in the representation of space correctly predict the structure of learned firing fields in trained neural networks. Further, our theory proves that a nonnegativity constraint on firing rates induces a symmetry-breaking mechanism which favors hexagonal firing fields. We extend this theory to the case of learning multiple grid maps and demonstrate that optimal solutions consist of a hierarchy of maps with increasing length scales. These results unify previous accounts of grid cell firing and provide a novel framework for predicting the learned representations of recurrent neural networks.
more »
« less
- Award ID(s):
- 1845166
- NSF-PAR ID:
- 10291261
- Date Published:
- Journal Name:
- Advances in neural information processing systems
- Volume:
- 32
- ISSN:
- 1049-5258
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Learned movements can be skillfully performed at different paces. What neural strategies produce this flexibility? Can they be predicted and understood by network modeling? We trained monkeys to perform a cycling task at different speeds, and trained artificial recurrent networks to generate the empirical muscle-activity patterns. Network solutions reflected the principle that smooth well-behaved dynamics require low trajectory tangling. Network solutions had a consistent form, which yielded quantitative and qualitative predictions. To evaluate predictions, we analyzed motor cortex activity recorded during the same task. Responses supported the hypothesis that the dominant neural signals reflect not muscle activity, but network-level strategies for generating muscle activity. Single-neuron responses were better accounted for by network activity than by muscle activity. Similarly, neural population trajectories shared their organization not with muscle trajectories, but with network solutions. Thus, cortical activity could be understood based on the need to generate muscle activity via dynamics that allow smooth, robust control over movement speed.more » « less
-
Feature representations from pre-trained deep neural networks have been known to exhibit excellent generalization and utility across a variety of related tasks. Fine-tuning is by far the simplest and most widely used approach that seeks to exploit and adapt these feature representations to novel tasks with limited data. Despite the effectiveness of fine-tuning, itis often sub-optimal and requires very careful optimization to prevent severe over-fitting to small datasets. The problem of sub-optimality and over-fitting, is due in part to the large number of parameters used in a typical deep convolutional neural network. To address these problems, we propose a simple yet effective regularization method for fine-tuning pre-trained deep networks for the task of k-shot learning. To prevent overfitting, our key strategy is to cluster the model parameters while ensuring intra-cluster similarity and inter-cluster diversity of the parameters, effectively regularizing the dimensionality of the parameter search space. In particular, we identify groups of neurons within each layer of a deep network that shares similar activation patterns. When the network is to be fine-tuned for a classification task using only k examples, we propagate a single gradient to all of the neuron parameters that belong to the same group. The grouping of neurons is non-trivial as neuron activations depend on the distribution of the input data. To efficiently search for optimal groupings conditioned on the input data, we propose a reinforcement learning search strategy using recurrent networks to learn the optimal group assignments for each network layer. Experimental results show that our method can be easily applied to several popular convolutional neural networks and improve upon other state-of-the-art fine-tuning based k-shot learning strategies by more than10%more » « less
-
more » « less
Abstract Motivation Computational methods for compound–protein affinity and contact (CPAC) prediction aim at facilitating rational drug discovery by simultaneous prediction of the strength and the pattern of compound–protein interactions. Although the desired outputs are highly structure-dependent, the lack of protein structures often makes structure-free methods rely on protein sequence inputs alone. The scarcity of compound–protein pairs with affinity and contact labels further limits the accuracy and the generalizability of CPAC models.
Results To overcome the aforementioned challenges of structure naivety and labeled-data scarcity, we introduce cross-modality and self-supervised learning, respectively, for structure-aware and task-relevant protein embedding. Specifically, protein data are available in both modalities of 1D amino-acid sequences and predicted 2D contact maps that are separately embedded with recurrent and graph neural networks, respectively, as well as jointly embedded with two cross-modality schemes. Furthermore, both protein modalities are pre-trained under various self-supervised learning strategies, by leveraging massive amount of unlabeled protein data. Our results indicate that individual protein modalities differ in their strengths of predicting affinities or contacts. Proper cross-modality protein embedding combined with self-supervised learning improves model generalizability when predicting both affinities and contacts for unseen proteins.
Availability and implementation Data and source codes are available at https://github.com/Shen-Lab/CPAC.
Supplementary information Supplementary data are available at Bioinformatics online.
-
null (Ed.)We investigate the ways in which a machine learning architecture known as Reservoir Computing learns concepts such as “similar” and “different” and other relationships between image pairs and generalizes these concepts to previously unseen classes of data. We present two Reservoir Computing architectures, which loosely resemble neural dynamics, and show that a Reservoir Computer (RC) trained to identify relationships between image pairs drawn from a subset of training classes generalizes the learned relationships to substantially different classes unseen during training. We demonstrate our results on the simple MNIST handwritten digit database as well as a database of depth maps of visual scenes in videos taken from a moving camera. We consider image pair relationships such as images from the same class; images from the same class with one image superposed with noise, rotated 90°, blurred, or scaled; images from different classes. We observe that the reservoir acts as a nonlinear filter projecting the input into a higher dimensional space in which the relationships are separable; i.e., the reservoir system state trajectories display different dynamical patterns that reflect the corresponding input pair relationships. Thus, as opposed to training in the entire high-dimensional reservoir space, the RC only needs to learns characteristic features of these dynamical patterns, allowing it to perform well with very few training examples compared with conventional machine learning feed-forward techniques such as deep learning. In generalization tasks, we observe that RCs perform significantly better than state-of-the-art, feed-forward, pair-based architectures such as convolutional and deep Siamese Neural Networks (SNNs). We also show that RCs can not only generalize relationships, but also generalize combinations of relationships, providing robust and effective image pair classification. Our work helps bridge the gap between explainable machine learning with small datasets and biologically inspired analogy-based learning, pointing to new directions in the investigation of learning processes.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
