Pankaj Dayama

Vijay Ekambaram, Arindam Jati, Pankaj Dayama et al.

Large pre-trained models excel in zero/few-shot learning for language and vision tasks but face challenges in multivariate time series (TS) forecasting due to diverse data characteristics. Consequently, recent research efforts have focused on developing pre-trained TS forecasting models. These models, whether built from scratch or adapted from large language models (LLMs), excel in zero/few-shot forecasting tasks. However, they are limited by slow performance, high computational demands, and neglect of cross-channel and exogenous correlations. To address this, we introduce Tiny Time Mixers (TTM), a compact model (starting from 1M parameters) with effective transfer learning capabilities, trained exclusively on public TS datasets. TTM, based on the light-weight TSMixer architecture, incorporates innovations like adaptive patching, diverse resolution sampling, and resolution prefix tuning to handle pre-training on varied dataset resolutions with minimal model capacity. Additionally, it employs multi-level modeling to capture channel correlations and infuse exogenous signals during fine-tuning. TTM outperforms existing popular benchmarks in zero/few-shot forecasting by (4-40%), while reducing computational requirements significantly. Moreover, TTMs are lightweight and can be executed even on CPU-only machines, enhancing usability and fostering wider adoption in resource-constrained environments. The model weights for reproducibility and research use are available at https://huggingface.co/ibm/ttm-research-r2/, while enterprise-use weights under the Apache license can be accessed as follows: the initial TTM-Q variant at https://huggingface.co/ibm-granite/granite-timeseries-ttm-r1, and the latest variants (TTM-B, TTM-E, TTM-A) weights are available at https://huggingface.co/ibm-granite/granite-timeseries-ttm-r2.

Vijay Ekambaram, Subodh Kumar, Arindam Jati et al.

The rise of time-series pre-trained models has advanced temporal representation learning, but current state-of-the-art models are often large-scale, requiring substantial compute. We introduce TSPulse, ultra-compact time-series pre-trained models with only 1M parameters, specialized to perform strongly across classification, anomaly detection, imputation, and retrieval tasks. TSPulse introduces innovations at both the architecture and task levels. At the architecture level, it employs a dual-space masked reconstruction, learning from both time and frequency domains to capture complementary signals. This is further enhanced by a dual-embedding disentanglement, generating both detailed embeddings for fine-grained analysis and high-level semantic embeddings for broader task understanding. Notably, TSPulse's semantic embeddings are robust to shifts in time, magnitude, and noise, which is important for robust retrieval. At the task level, TSPulse incorporates TSLens, a fine-tuning component enabling task-specific feature attention. It also introduces a multi-head triangulation technique that correlates deviations from multiple prediction heads, enhancing anomaly detection by fusing complementary model outputs. Additionally, a hybrid mask pretraining is proposed to improves zero-shot imputation by reducing pre-training bias. These architecture and task innovations collectively contribute to TSPulse's significant performance gains: 5-16% on the UEA classification benchmarks, +20% on the TSB-AD anomaly detection leaderboard, +50% in zero-shot imputation, and +25% in time-series retrieval. Remarkably, these results are achieved with just 1M parameters (10-100X smaller than existing SOTA models) and allow GPU-free inference, setting a new standard for efficient time-series pre-trained models. The models can be accessed from https://huggingface.co/ibm-granite/granite-timeseries-tspulse-r1

Chandramouli Kamanchi, Sumanta Mukherjee, Kameshwaran Sampath et al.

Activation functions are non-linearities in neural networks that allow them to learn complex mapping between inputs and outputs. Typical choices for activation functions are ReLU, Tanh, Sigmoid etc., where the choice generally depends on the application domain. In this work, we propose a framework/strategy that unifies several works on activation functions and theoretically explains the performance benefits of these works. We also propose novel techniques that originate from the framework and allow us to obtain ``extensions'' (i.e. special generalizations of a given neural network) of neural networks through operations on activation functions. We theoretically and empirically show that ``extensions'' of neural networks have performance benefits compared to vanilla neural networks with insignificant space and time complexity costs on standard test functions. We also show the benefits of neural network ``extensions'' in the time-series domain on real-world datasets.

Sneha Chaudhari, Pankaj Dayama, Vinayaka Pandit et al.

In this paper, we consider active information acquisition when the prediction model is meant to be applied on a targeted subset of the population. The goal is to label a pre-specified fraction of customers in the target or test set by iteratively querying for information from the non-target or training set. The number of queries is limited by an overall budget. Arising in the context of two rather disparate applications- banking and medical diagnosis, we pose the active information acquisition problem as a constrained optimization problem. We propose two greedy iterative algorithms for solving the above problem. We conduct experiments with synthetic data and compare results of our proposed algorithms with few other baseline approaches. The experimental results show that our proposed approaches perform better than the baseline schemes.