Luca Herranz-Celotti

Luca Herranz-Celotti, Jean Rouat

Numerous theories of learning propose to prevent the gradient from exponential growth with depth or time, to stabilize and improve training. Typically, these analyses are conducted on feed-forward fully-connected neural networks or simple single-layer recurrent neural networks, given their mathematical tractability. In contrast, this study demonstrates that pre-training the network to local stability can be effective whenever the architectures are too complex for an analytical initialization. Furthermore, we extend known stability theories to encompass a broader family of deep recurrent networks, requiring minimal assumptions on data and parameter distribution, a theory we call the Local Stability Condition (LSC). Our investigation reveals that the classical Glorot, He, and Orthogonal initialization schemes satisfy the LSC when applied to feed-forward fully-connected neural networks. However, analysing deep recurrent networks, we identify a new additive source of exponential explosion that emerges from counting gradient paths in a rectangular grid in depth and time. We propose a new approach to mitigate this issue, that consists on giving a weight of a half to the time and depth contributions to the gradient, instead of the classical weight of one. Our empirical results confirm that pre-training both feed-forward and recurrent networks, for differentiable, neuromorphic and state-space models to fulfill the LSC, often results in improved final performance. This study contributes to the field by providing a means to stabilize networks of any complexity. Our approach can be implemented as an additional step before pre-training on large augmented datasets, and as an alternative to finding stable initializations analytically.

Luca Herranz-Celotti, Ermal Rrapaj

The softmax attention mechanism has emerged as a noteworthy development in the field of Artificial Intelligence research, building on the successes of Transformer-based architectures. However, their ever increasing sizes necessitate ever increasing computational memory, that limits their usage. We propose KgV, a sigmoid gating mechanism that, in conjunction with softmax attention, significantly boosts performance without increasing architecture size. To amend the size requirements, we leverage Tensor Chains to identify and prune the excess parameters. We find that such excess resides primarily within the embedding layer, and not in the output linear layer. To further improve embedding and significantly reduce parameters, we introduce H-SoftPOS, a hierarchical embedding layer which simultaneously enhances performance. Remarkably, on the WMT14 English-German validation set, our approach yields a threefold reduction in perplexity, surpassing the current state-of-the-art, while reducing parameter counts also by a factor of 3. When we further reduce the number of parameters up to sevenfold, we can still achieve a 21\% decrease in perplexity with respect to the baseline Transformer. To understand generalization capabilities, we conduct experiments on the 7 language pairs of the WMT17 dataset. Our method outperforms existing techniques in terms of test loss while simultaneously halving the number of parameters. Moreover, we observe a 70 times reduction in variance with respect to the prior state-of-the-art. In conclusion, our proposed method yields significant improvements in performance and much lower memory cost. We call the resulting architecture Anthe.

Luca Herranz-Celotti, Jean Rouat

Stability arguments are often used to prevent learning algorithms from having ever increasing activity and weights that hinder generalization. However, stability conditions can clash with the sparsity required to augment the energy efficiency of spiking neurons. Nonetheless it can also provide solutions. In fact, spiking Neuromorphic Computing uses binary activity to improve Artificial Intelligence energy efficiency. However, its non-smoothness requires approximate gradients, known as Surrogate Gradients (SG), to close the performance gap with Deep Learning. Several SG have been proposed in the literature, but it remains unclear how to determine the best SG for a given task and network. Thus, we aim at theoretically define the best SG, through stability arguments, to reduce the need for grid search. In fact, we show that more complex tasks and networks need more careful choice of SG, even if overall the derivative of the fast sigmoid tends to outperform the other, for a wide range of learning rates. We therefore design a stability based theoretical method to choose initialization and SG shape before training on the most common spiking neuron, the Leaky Integrate and Fire (LIF). Since our stability method suggests the use of high firing rates at initialization, which is non-standard in the neuromorphic literature, we show that high initial firing rates, combined with a sparsity encouraging loss term introduced gradually, can lead to better generalization, depending on the SG shape. Our stability based theoretical solution, finds a SG and initialization that experimentally result in improved accuracy. We show how it can be used to reduce the need of extensive grid-search of dampening, sharpness and tail-fatness of the SG. We also show that our stability concepts can be extended to be applicable on different LIF variants, such as DECOLLE and fluctuations-driven initializations.