Towards an integration of deep learning and neuroscience
arXiv:1606.03813 · doi:10.3389/fncom.2016.00094
Abstract
Neuroscience has focused on the detailed implementation of computation, studying neural codes, dynamics and circuits. In machine learning, however, artificial neural networks tend to eschew precisely designed codes, dynamics or circuits in favor of brute force optimization of a cost function, often using simple and relatively uniform initial architectures. Two recent developments have emerged within machine learning that create an opportunity to connect these seemingly divergent perspectives. First, structured architectures are used, including dedicated systems for attention, recursion and various forms of short- and long-term memory storage. Second, cost functions and training procedures have become more complex and are varied across layers and over time. Here we think about the brain in terms of these ideas. We hypothesize that (1) the brain optimizes cost functions, (2) these cost functions are diverse and differ across brain locations and over development, and (3) optimization operates within a pre-structured architecture matched to the computational problems posed by behavior. Such a heterogeneously optimized system, enabled by a series of interacting cost functions, serves to make learning data-efficient and precisely targeted to the needs of the organism. We suggest directions by which neuroscience could seek to refine and test these hypotheses.
References in corpus (29)
- Deep Learning in Neural Networks: An Overview
- Pixel Recurrent Neural Networks
- Deep Neural Networks Reveal a Gradient in the Complexity of Neural Representations across the Brain's Ventral Visual Pathway
- Exact solutions to the nonlinear dynamics of learning in deep linear neural networks
- Why Neurons Have Thousands of Synapses, A Theory of Sequence Memory in Neocortex
- Recurrent Network Models Of Sequence Generation And Memory
- Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning
- Towards Biologically Plausible Deep Learning
- Attend, Infer, Repeat: Fast Scene Understanding with Generative Models
- Bridging the Gaps Between Residual Learning, Recurrent Neural Networks and Visual Cortex
- Network Plasticity as Bayesian Inference
- Foveation-based Mechanisms Alleviate Adversarial Examples
- Knowledge Matters: Importance of Prior Information for Optimization
- Gradient learning in spiking neural networks by dynamic perturbation of conductances
- How Auto-Encoders Could Provide Credit Assignment in Deep Networks via Target Propagation
- Learning to Compose Neural Networks for Question Answering
- Learning the Pseudoinverse Solution to Network Weights
- Frequently Asked Questions for: The Atoms of Neural Computation
- Random feedback weights support learning in deep neural networks
- Unsupervised Learning of Visual Structure using Predictive Generative Networks
- Nonlinear Hebbian learning as a unifying principle in receptive field formation
- Learning Through Time in the Thalamocortical Loops
- An Algorithm for Training Polynomial Networks
- STDP as presynaptic activity times rate of change of postsynaptic activity
- Learning with hidden variables
- Early Inference in Energy-Based Models Approximates Back-Propagation
- The grammar of mammalian brain capacity
- Cortical prediction markets
- Recurrent Exponential-Family Harmoniums without Backprop-Through-Time