Update paper.md

JOSS/paper.md

@@ -37,11 +37,11 @@ Designed to be lightweight and user-friendly, PyBCI emphasizes quick customizati
 
 # State of the Field:
 
-There are a variety of BCI software packages available, each with its own advantages and limitations. Notable packages include solutions like OpenViBE[@OpenViBE] and BCI2000[@BCI2000] offer ease of use for those without programming expertise. BciPy[@BciPy], another Python-based platform, provides some level of customization but does not allow for the easy integration of popular machine learning libraries. In contrast, PyBCI offers seamless integration with a variety of machine learning libraries and feature extraction tools. This flexibility makes PyBCI a robust choice for researchers seeking a tailored, code-based approach to their BCI experiments.
+There are a variety of BCI software packages available, each with its own advantages and limitations. Notable packages include solutions like OpenViBE[@OpenViBE] and BCI2000[@BCI2000] that offer ease of use for those without programming expertise. BciPy[@BciPy], another Python-based platform, provides some level of customization but does not allow for the easy integration of popular machine learning libraries. In contrast, PyBCI offers seamless integration with a variety of machine learning libraries and feature extraction tools. This flexibility makes PyBCI a robust choice for researchers seeking a tailored, code-based approach to their BCI experiments.
 
 # Software functionality and performance:
 
-PyBCI accelerates the pace of BCI research by streamlining data collection, processing, and model analysis. It uses the Lab Streaming Layer (LSL) to handle data acquisition and labelling, allowing for real-time, synchronous data collection from multiple devices [@lsl]. Samples are collected in chunks from the LSL data streams and stored in pre-allocated NumPy arrays. When in training mode based on a configurable time window before and after each marker type. When in test mode data is continuously processed and analysed based on the global epoch timing settings.  For feature extraction, PyBCI leverages the power of NumPy [@oliphant2006guide], SciPy [@2020SciPy-NMeth], and Antropy [@vallat_antropy_2023] robust Python libraries known for their efficiency in handling numerical operations. Machine learning, a crucial component of BCI research, is facilitated with PyTorch [@NEURIPS2019_9015], SciKit-learn [@scikit-learn] and TensorFlow [@tensorflow2015-whitepaper]. Scikit-learn offers a wide range of traditional algorithms for classification, including things like regression, and clustering, while TensorFlow and PyTorch provide comprehensive ecosystems for developing and training bespoke deep learning machine learning models.
+PyBCI accelerates the pace of BCI research by streamlining data collection, processing, and model analysis. It uses the Lab Streaming Layer (LSL) to handle data acquisition and labelling, allowing for real-time, synchronous data collection from multiple devices [@lsl]. Samples are collected in chunks from the LSL data streams and stored in pre-allocated NumPy arrays. When in training mode based on a configurable time window before and after each marker type. When in test mode, data is continuously processed and analysed based on the global epoch timing settings.  For feature extraction, PyBCI leverages the power of NumPy [@oliphant2006guide], SciPy [@2020SciPy-NMeth], and Antropy [@vallat_antropy_2023], robust Python libraries known for their efficiency in handling numerical operations. Machine learning, a crucial component of BCI research, is facilitated with PyTorch [@NEURIPS2019_9015], SciKit-learn [@scikit-learn] and TensorFlow [@tensorflow2015-whitepaper]. Scikit-learn offers a wide range of traditional algorithms for classification, including things like regression, and clustering, while TensorFlow and PyTorch provide comprehensive ecosystems for developing and training bespoke deep learning machine learning models.
 
 # Impact: