Biologically plausible unsupervised learning in shallow and deep neural networks

The way our brain learns to disentangle complex signals into unambiguous concepts is fascinating but remains largely unknown. There is evidence, however, that hierarchical neural representations play a key role in the cortex. This thesis investigates biologically plausible models of unsupervised learning of hierarchical representations as found in the brain and modern computer vision models. We use computational modeling to address three main questions at the intersection of artificial intelligence (AI) and computational neuroscience.The first question is: What are useful neural representations and when are deep hierarchical representations needed? We approach this point with a systematic study of biologically plausible unsupervised feature learning in a shallow 2-layer networks on digit (MNIST) and object (CIFAR10) classification. Surprisingly, random features support high performance, especially for large hidden layers. When combined with localized receptive fields, random feature networks approach the performance of supervised backpropagation on MNIST, but not on CIFAR10. We suggest that future models of biologically plausible learning should outperform such random feature benchmarks on MNIST, or that such models should be evaluated in different ways.The second question is: How can hierarchical representations be learned with mechanisms supported by neuroscientific evidence? We cover this question by proposing a unifying Hebbian model, inspired by common models of V1 simple and complex cells based on unsupervised sparse coding and temporal invariance learning. In shallow 2-layer networks, our model reproduces learning of simple and complex cell receptive fields, as found in V1. In deeper networks, we stack multiple layers of Hebbian learning but find that it does not yield hierarchical representations of increasing usefulness. From this, we hypothesise that standard Hebbian rules are too constrained to build increasingly useful representations, as observed in higher areas of the visual cortex or deep artificial neural networks.The third question is: Can AI inspire learning models that build deep representations and are still biologically plausible? We address this question by proposing a learning rule that takes inspiration from neuroscience and recent advances in self-supervised deep learning. The proposed rule is Hebbian, i.e. only depends on pre- and post-synaptic neuronal activity, but includes additional local factors, namely predictive dendritic input and widely broadcasted modulation factors. Algorithmically, this rule applies self-supervised contrastive predictive learning to a causal, biological setting using saccades. We find that networks trained with this generalised Hebbian rule build deep hierarchical representations of images, speech and video.We see our modeling as a potential starting point for both, new hypotheses, that can be tested experimentally, and novel AI models that could benefit from added biological realism.

Theory of representation learning in cortical neural networks

Carlos Stein Naves de Brito

Our brain continuously self-organizes to construct and maintain an internal representation of the world based on the information arriving through sensory stimuli. Remarkably, cortical areas related to different sensory modalities appear to share the same functional unit, the neuron, and develop through the same learning mechanism, synaptic plasticity. It motivates the conjecture of a unifying theory to explain cortical representational learning across sensory modalities. In this thesis we present theories and computational models of learning and optimization in neural networks, postulating functional properties of synaptic plasticity that support the apparent universal learning capacity of cortical networks. In the past decades, a variety of theories and models have been proposed to describe receptive field formation in sensory areas. They include normative models such as sparse coding, and bottom-up models such as spike-timing dependent plasticity. We bring together candidate explanations by demonstrating that in fact a single principle is sufficient to explain receptive field development. First, we show that many representative models of sensory development are in fact implementing variations of a common principle: nonlinear Hebbian learning. Second, we reveal that nonlinear Hebbian learning is sufficient for receptive field formation through sensory inputs. A surprising result is that our findings are independent of specific details, and allow for robust predictions of the learned receptive fields. Thus nonlinear Hebbian learning and natural statistics can account for many aspects of receptive field formation across models and sensory modalities. The Hebbian learning theory substantiates that synaptic plasticity can be interpreted as an optimization procedure, implementing stochastic gradient descent. In stochastic gradient descent inputs arrive sequentially, as in sensory streams. However, individual data samples have very little information about the correct learning signal, and it becomes a fundamental problem to know how many samples are required for reliable synaptic changes. Through estimation theory, we develop a novel adaptive learning rate model, that adapts the magnitude of synaptic changes based on the statistics of the learning signal, enabling an optimal use of data samples. Our model has a simple implementation and demonstrates improved learning speed, making this a promising candidate for large artificial neural network applications. The model also makes predictions on how cortical plasticity may modulate synaptic plasticity for optimal learning. The optimal sampling size for reliable learning allows us to estimate optimal learning times for a given model. We apply this theory to derive analytical bounds on times for the optimization of synaptic connections. First, we show this optimization problem to have exponentially many saddle-nodes, which lead to small gradients and slow learning. Second, we show that the number of input synapses to a neuron modulates the magnitude of the initial gradient, determining the duration of learning. Our final result reveals that the learning duration increases supra-linearly with the number of synapses, suggesting an effective limit on synaptic connections and receptive field sizes in developing neural networks.

EPFL2016

Enactive robot vision

Mototaka Suzuki

The complexity of today's autonomous robots poses a major challenge for Artificial Intelligence. These robots are equipped with sophisticated sensors and mechanical abilities that allow them to enter our homes and interact with humans. For example, today's robots are almost all equipped with vision and several of them can move over rough terrain with wheels or legs. The methods developed so far in Artificial Intelligence, however, are not yet ready to cope with the complexity of the information gathered through the robot sensors and the need for rapid action in partially unknown and dynamic environments. In this thesis, I will argue that the apparent complexity of the environment and of the robot brain can be significantly simplified if perception, behavior, and learning are allowed to co-develop on the same time scale. In doing so, robots become sensitive to, and actively exploit, characteristics of the environment that they can tackle within their own computational and physical constraints. This line of work is grounded on philosophical and psychological research showing that perception is an active process mediated by behavior. However, computational models of active vision are very rare and often rely on architectures that are preprogrammed to detect certain characteristics of the environment. Previous work have shown that complex visual tasks, such as position and size invariant shape recognition as well as navigation, can be tackled with remarkably simple neural architectures generated by a coevolutionary process of active vision and feature selection. Behavioral machines equipped with primitive vision systems and direct pathways between visual and motor neurons were evolved while freely interacting with their environments. I proceed further on this line of investigation and describe the application of this methodology in three situations, namely car driving with an omnidirectional camera, goal-oriented navigation of a humanoid robot, and cooperative tasks by two agents. I will show that these systems develop sensitivity to a number of oriented, retinotopic, visual features – oriented edges, corners, height – and a behavioral repertoire to locate, bring, and keep these features in sensitive regions of the vision system that allow them to accomplish their goals. In a second set of experiments, I will show that active vision can be exploited by the robot to perform anticipatory exploration of the environment in a task that requires landmark-based navigation. Evolved robots exploit an internal expectation system that makes use of active exploration to check for expected events in the environment. I will describe a third set of experiments where, in addition to an evolutionary process, the visual system of the robot can develop receptive fields by means of unsupervised Hebbian learning and show that these receptive fields are significantly affected by the behavior of the system and differ from those predicted by most computational models of visual cortex. Finally, I will show that these robots replicate the performance deficiencies observed in experiments of motor deprivation with kitten when they are exposed to the same type of motor deprivations. Furthermore, the analyses of our robot brains suggest an explanation for the deficiencies observed in kitten that have not yet been fully understood.

EPFL2007

Lifelong Machine Learning with Data Efficiency and Knowledge Retention

Fei Mi

Artificial intelligence (AI) and machine learning (ML) have become de facto tools in many real-life applications to offer a wide range of benefits for individuals and our society. A classic ML model is typically trained with a large-scale static dataset in an offline manner. Therefore, it can not quickly capture new knowledge in non-stationary environments, and it is difficult to maintain long-term memory for knowledge learned earlier. In practice, many ML systems often need to learn new knowledge (e.g., domains, tasks, distributions, etc.) as more data and experiences are collected, which is referred to as a lifelong ML paradigm in this thesis. We focus on two fundamental challenges to achieve lifelong learning. The first challenge is to quickly learn new knowledge with a small number of observations, and we refer to it as data efficiency. The second challenge is to prevent an ML system from forgetting the old knowledge it has previously learned, and we refer to this challenge as knowledge retention. These two capabilities are crucial for applying ML to most practical applications. In this thesis, we study three important applications with these two challenges, including recommendation systems, task-oriented dialog systems, and the image classification task. First, we propose two approaches to improve data efficiency for task-oriented dialog systems. The first proposed approach is based on Meta-learning, aiming to learn a better model parameter initialization from training data. It can quickly reach a good parameter region of new domains or tasks with a small number of labeled data. The second proposal takes a semi-supervised self-training approach to iteratively train a better model using sufficient unlabeled data when only a limited number of labeled data are available. We empirically demonstrate that both approaches effectively improve data efficiency to learn new knowledge. The second self-training method even consistently improves state-of-the-art large-scale pre-trained models. Second, we tackle the knowledge retention challenge to mitigate the detrimental catastrophic forgetting issue when neural networks learn new knowledge sequentially. We formulate and investigate the ``continual learning'' setting for task-oriented dialog systems and recommendation systems. Through extensive empirical evaluation and analysis, we demonstrate the importance of (1) exemplar replay: storing representative historical data and replaying them to the model while learning new knowledge; (2) dynamic regularization: applying a dynamic regularization term to put flexible constraints on not forgetting previously learned knowledge in each model update cycle. Lastly, we conduct several initial attempts to achieve both data efficiency and knowledge retention in a unified framework. In the recommendation scenario, we propose two approaches using different non-parametric memory modules to retain long-term knowledge. More importantly, the two proposed non-parametric predictions computed on top of them help learn and memorize new knowledge in a data-efficient manner. Apart from the recommendation scenario, we propose a probabilistic evaluation protocol in the widely studied image classification domain. It is general and versatile to simulate a wide range of realistic lifelong learning scenarios that require both knowledge retention and data efficiency for studying different techniques. Through experiments, we also demonstrate the benefit

EPFL2021